API Reference

mmdet.apis

mmdet.apis.get_root_logger(log_file=None, log_level=20)[source]

Get root logger.

Parameters:
  • log_file (str, optional) – File path of log. Defaults to None.
  • log_level (int, optional) – The level of logger. Defaults to logging.INFO.
Returns:

The obtained logger

Return type:

logging.Logger

mmdet.apis.set_random_seed(seed, deterministic=False)[source]

Set random seed.

Parameters:
  • seed (int) – Seed to be used.
  • deterministic (bool) – Whether to set the deterministic option for CUDNN backend, i.e., set torch.backends.cudnn.deterministic to True and torch.backends.cudnn.benchmark to False. Default: False.
mmdet.apis.init_detector(config, checkpoint=None, device='cuda:0', cfg_options=None)[source]

Initialize a detector from config file.

Parameters:
  • config (str or mmcv.Config) – Config file path or the config object.
  • checkpoint (str, optional) – Checkpoint path. If left as None, the model will not load any weights.
  • cfg_options (dict) – Options to override some settings in the used config.
Returns:

The constructed detector.

Return type:

nn.Module

mmdet.apis.async_inference_detector(model, imgs)[source]

Async inference image(s) with the detector.

Parameters:
  • model (nn.Module) – The loaded detector.
  • img (str | ndarray) – Either image files or loaded images.
Returns:

Awaitable detection results.

mmdet.apis.inference_detector(model, imgs)[source]

Inference image(s) with the detector.

Parameters:
  • model (nn.Module) – The loaded detector.
  • imgs (str/ndarray or list[str/ndarray] or tuple[str/ndarray]) – Either image files or loaded images.
Returns:

If imgs is a list or tuple, the same length list type results will be returned, otherwise return the detection results directly.

mmdet.apis.show_result_pyplot(model, img, result, score_thr=0.3, title='result', wait_time=0)[source]

Visualize the detection results on the image.

Parameters:
  • model (nn.Module) – The loaded detector.
  • img (str or np.ndarray) – Image filename or loaded image.
  • result (tuple[list] or list) – The detection result, can be either (bbox, segm) or just bbox.
  • score_thr (float) – The threshold to visualize the bboxes and masks.
  • title (str) – Title of the pyplot figure.
  • wait_time (float) – Value of waitKey param. Default: 0.
mmdet.apis.multi_gpu_test(model, data_loader, tmpdir=None, gpu_collect=False)[source]

Test model with multiple gpus.

This method tests model with multiple gpus and collects the results under two different modes: gpu and cpu modes. By setting ‘gpu_collect=True’ it encodes results to gpu tensors and use gpu communication for results collection. On cpu mode it saves the results on different gpus to ‘tmpdir’ and collects them by the rank 0 worker.

Parameters:
  • model (nn.Module) – Model to be tested.
  • data_loader (nn.Dataloader) – Pytorch data loader.
  • tmpdir (str) – Path of directory to save the temporary results from different gpus under cpu mode.
  • gpu_collect (bool) – Option to use either gpu or cpu to collect results.
Returns:

The prediction results.

Return type:

list

mmdet.core

anchor

class mmdet.core.anchor.AnchorGenerator(strides, ratios, scales=None, base_sizes=None, scale_major=True, octave_base_scale=None, scales_per_octave=None, centers=None, center_offset=0.0)[source]

Standard anchor generator for 2D anchor-based detectors.

Parameters:
  • strides (list[int] | list[tuple[int, int]]) – Strides of anchors in multiple feature levels in order (w, h).
  • ratios (list[float]) – The list of ratios between the height and width of anchors in a single level.
  • scales (list[int] | None) – Anchor scales for anchors in a single level. It cannot be set at the same time if octave_base_scale and scales_per_octave are set.
  • base_sizes (list[int] | None) – The basic sizes of anchors in multiple levels. If None is given, strides will be used as base_sizes. (If strides are non square, the shortest stride is taken.)
  • scale_major (bool) – Whether to multiply scales first when generating base anchors. If true, the anchors in the same row will have the same scales. By default it is True in V2.0
  • octave_base_scale (int) – The base scale of octave.
  • scales_per_octave (int) – Number of scales for each octave. octave_base_scale and scales_per_octave are usually used in retinanet and the scales should be None when they are set.
  • centers (list[tuple[float, float]] | None) – The centers of the anchor relative to the feature grid center in multiple feature levels. By default it is set to be None and not used. If a list of tuple of float is given, they will be used to shift the centers of anchors.
  • center_offset (float) – The offset of center in proportion to anchors’ width and height. By default it is 0 in V2.0.

Examples

>>> from mmdet.core import AnchorGenerator
>>> self = AnchorGenerator([16], [1.], [1.], [9])
>>> all_anchors = self.grid_anchors([(2, 2)], device='cpu')
>>> print(all_anchors)
[tensor([[-4.5000, -4.5000,  4.5000,  4.5000],
        [11.5000, -4.5000, 20.5000,  4.5000],
        [-4.5000, 11.5000,  4.5000, 20.5000],
        [11.5000, 11.5000, 20.5000, 20.5000]])]
>>> self = AnchorGenerator([16, 32], [1.], [1.], [9, 18])
>>> all_anchors = self.grid_anchors([(2, 2), (1, 1)], device='cpu')
>>> print(all_anchors)
[tensor([[-4.5000, -4.5000,  4.5000,  4.5000],
        [11.5000, -4.5000, 20.5000,  4.5000],
        [-4.5000, 11.5000,  4.5000, 20.5000],
        [11.5000, 11.5000, 20.5000, 20.5000]]),         tensor([[-9., -9., 9., 9.]])]
gen_base_anchors()[source]

Generate base anchors.

Returns:Base anchors of a feature grid in multiple feature levels.
Return type:list(torch.Tensor)
gen_single_level_base_anchors(base_size, scales, ratios, center=None)[source]

Generate base anchors of a single level.

Parameters:
  • base_size (int | float) – Basic size of an anchor.
  • scales (torch.Tensor) – Scales of the anchor.
  • ratios (torch.Tensor) – The ratio between between the height and width of anchors in a single level.
  • center (tuple[float], optional) – The center of the base anchor related to a single feature grid. Defaults to None.
Returns:

Anchors in a single-level feature maps.

Return type:

torch.Tensor

grid_anchors(featmap_sizes, device='cuda')[source]

Generate grid anchors in multiple feature levels.

Parameters:
  • featmap_sizes (list[tuple]) – List of feature map sizes in multiple feature levels.
  • device (str) – Device where the anchors will be put on.
Returns:

Anchors in multiple feature levels. The sizes of each tensor should be [N, 4], where N = width * height * num_base_anchors, width and height are the sizes of the corresponding feature level, num_base_anchors is the number of anchors for that level.

Return type:

list[torch.Tensor]

grid_priors(featmap_sizes, device='cuda')[source]

Generate grid anchors in multiple feature levels.

Parameters:
  • featmap_sizes (list[tuple]) – List of feature map sizes in multiple feature levels.
  • device (str) – The device where the anchors will be put on.
Returns:

Anchors in multiple feature levels. The sizes of each tensor should be [N, 4], where N = width * height * num_base_anchors, width and height are the sizes of the corresponding feature level, num_base_anchors is the number of anchors for that level.

Return type:

list[torch.Tensor]

num_base_anchors

total number of base anchors in a feature grid

Type:list[int]
num_base_priors

The number of priors (anchors) at a point on the feature grid

Type:list[int]
num_levels

number of feature levels that the generator will be applied

Type:int
single_level_grid_anchors(base_anchors, featmap_size, stride=(16, 16), device='cuda')[source]

Generate grid anchors of a single level.

Note

This function is usually called by method self.grid_anchors.

Parameters:
  • base_anchors (torch.Tensor) – The base anchors of a feature grid.
  • featmap_size (tuple[int]) – Size of the feature maps.
  • stride (tuple[int], optional) – Stride of the feature map in order (w, h). Defaults to (16, 16).
  • device (str, optional) – Device the tensor will be put on. Defaults to ‘cuda’.
Returns:

Anchors in the overall feature maps.

Return type:

torch.Tensor

single_level_grid_priors(featmap_size, level_idx, device='cuda')[source]

Generate grid anchors of a single level.

Note

This function is usually called by method self.grid_priors.

Parameters:
  • featmap_size (tuple[int]) – Size of the feature maps.
  • level_idx (int) – The index of corresponding feature map level.
  • device (str, optional) – The device the tensor will be put on. Defaults to ‘cuda’.
Returns:

Anchors in the overall feature maps.

Return type:

torch.Tensor

single_level_valid_flags(featmap_size, valid_size, num_base_anchors, device='cuda')[source]

Generate the valid flags of anchor in a single feature map.

Parameters:
  • featmap_size (tuple[int]) – The size of feature maps, arrange as (h, w).
  • valid_size (tuple[int]) – The valid size of the feature maps.
  • num_base_anchors (int) – The number of base anchors.
  • device (str, optional) – Device where the flags will be put on. Defaults to ‘cuda’.
Returns:

The valid flags of each anchor in a single level feature map.

Return type:

torch.Tensor

sparse_priors(prior_idxs, featmap_size, level_idx, dtype=torch.float32, device='cuda')[source]

Generate sparse anchors according to the prior_idxs.

Parameters:
  • prior_idxs (Tensor) – The index of corresponding anchors in the feature map.
  • featmap_size (tuple[int]) – feature map size arrange as (h, w).
  • level_idx (int) – The level index of corresponding feature map.
  • (obj (device) – torch.dtype): Date type of points.Defaults to torch.float32.
  • (objtorch.device): The device where the points is located.
Returns:

Anchor with shape (N, 4), N should be equal to

the length of prior_idxs.

Return type:

Tensor

valid_flags(featmap_sizes, pad_shape, device='cuda')[source]

Generate valid flags of anchors in multiple feature levels.

Parameters:
  • featmap_sizes (list(tuple)) – List of feature map sizes in multiple feature levels.
  • pad_shape (tuple) – The padded shape of the image.
  • device (str) – Device where the anchors will be put on.
Returns:

Valid flags of anchors in multiple levels.

Return type:

list(torch.Tensor)

class mmdet.core.anchor.LegacyAnchorGenerator(strides, ratios, scales=None, base_sizes=None, scale_major=True, octave_base_scale=None, scales_per_octave=None, centers=None, center_offset=0.0)[source]

Legacy anchor generator used in MMDetection V1.x.

Note

Difference to the V2.0 anchor generator:

  1. The center offset of V1.x anchors are set to be 0.5 rather than 0.
  2. The width/height are minused by 1 when calculating the anchors’ centers and corners to meet the V1.x coordinate system.
  3. The anchors’ corners are quantized.
Parameters:
  • strides (list[int] | list[tuple[int]]) – Strides of anchors in multiple feature levels.
  • ratios (list[float]) – The list of ratios between the height and width of anchors in a single level.
  • scales (list[int] | None) – Anchor scales for anchors in a single level. It cannot be set at the same time if octave_base_scale and scales_per_octave are set.
  • base_sizes (list[int]) – The basic sizes of anchors in multiple levels. If None is given, strides will be used to generate base_sizes.
  • scale_major (bool) – Whether to multiply scales first when generating base anchors. If true, the anchors in the same row will have the same scales. By default it is True in V2.0
  • octave_base_scale (int) – The base scale of octave.
  • scales_per_octave (int) – Number of scales for each octave. octave_base_scale and scales_per_octave are usually used in retinanet and the scales should be None when they are set.
  • centers (list[tuple[float, float]] | None) – The centers of the anchor relative to the feature grid center in multiple feature levels. By default it is set to be None and not used. It a list of float is given, this list will be used to shift the centers of anchors.
  • center_offset (float) – The offset of center in propotion to anchors’ width and height. By default it is 0.5 in V2.0 but it should be 0.5 in v1.x models.

Examples

>>> from mmdet.core import LegacyAnchorGenerator
>>> self = LegacyAnchorGenerator(
>>>     [16], [1.], [1.], [9], center_offset=0.5)
>>> all_anchors = self.grid_anchors(((2, 2),), device='cpu')
>>> print(all_anchors)
[tensor([[ 0.,  0.,  8.,  8.],
        [16.,  0., 24.,  8.],
        [ 0., 16.,  8., 24.],
        [16., 16., 24., 24.]])]
gen_single_level_base_anchors(base_size, scales, ratios, center=None)[source]

Generate base anchors of a single level.

Note

The width/height of anchors are minused by 1 when calculating the centers and corners to meet the V1.x coordinate system.

Parameters:
  • base_size (int | float) – Basic size of an anchor.
  • scales (torch.Tensor) – Scales of the anchor.
  • ratios (torch.Tensor) – The ratio between between the height. and width of anchors in a single level.
  • center (tuple[float], optional) – The center of the base anchor related to a single feature grid. Defaults to None.
Returns:

Anchors in a single-level feature map.

Return type:

torch.Tensor

mmdet.core.anchor.anchor_inside_flags(flat_anchors, valid_flags, img_shape, allowed_border=0)[source]

Check whether the anchors are inside the border.

Parameters:
  • flat_anchors (torch.Tensor) – Flatten anchors, shape (n, 4).
  • valid_flags (torch.Tensor) – An existing valid flags of anchors.
  • img_shape (tuple(int)) – Shape of current image.
  • allowed_border (int, optional) – The border to allow the valid anchor. Defaults to 0.
Returns:

Flags indicating whether the anchors are inside a valid range.

Return type:

torch.Tensor

mmdet.core.anchor.images_to_levels(target, num_levels)[source]

Convert targets by image to targets by feature level.

[target_img0, target_img1] -> [target_level0, target_level1, …]

mmdet.core.anchor.calc_region(bbox, ratio, featmap_size=None)[source]

Calculate a proportional bbox region.

The bbox center are fixed and the new h’ and w’ is h * ratio and w * ratio.

Parameters:
  • bbox (Tensor) – Bboxes to calculate regions, shape (n, 4).
  • ratio (float) – Ratio of the output region.
  • featmap_size (tuple) – Feature map size used for clipping the boundary.
Returns:

x1, y1, x2, y2

Return type:

tuple

class mmdet.core.anchor.YOLOAnchorGenerator(strides, base_sizes)[source]

Anchor generator for YOLO.

Parameters:
  • strides (list[int] | list[tuple[int, int]]) – Strides of anchors in multiple feature levels.
  • base_sizes (list[list[tuple[int, int]]]) – The basic sizes of anchors in multiple levels.
gen_base_anchors()[source]

Generate base anchors.

Returns:Base anchors of a feature grid in multiple feature levels.
Return type:list(torch.Tensor)
gen_single_level_base_anchors(base_sizes_per_level, center=None)[source]

Generate base anchors of a single level.

Parameters:
  • base_sizes_per_level (list[tuple[int, int]]) – Basic sizes of anchors.
  • center (tuple[float], optional) – The center of the base anchor related to a single feature grid. Defaults to None.
Returns:

Anchors in a single-level feature maps.

Return type:

torch.Tensor

num_levels

number of feature levels that the generator will be applied

Type:int
responsible_flags(featmap_sizes, gt_bboxes, device='cuda')[source]

Generate responsible anchor flags of grid cells in multiple scales.

Parameters:
  • featmap_sizes (list(tuple)) – List of feature map sizes in multiple feature levels.
  • gt_bboxes (Tensor) – Ground truth boxes, shape (n, 4).
  • device (str) – Device where the anchors will be put on.
Returns:

responsible flags of anchors in multiple level

Return type:

list(torch.Tensor)

single_level_responsible_flags(featmap_size, gt_bboxes, stride, num_base_anchors, device='cuda')[source]

Generate the responsible flags of anchor in a single feature map.

Parameters:
  • featmap_size (tuple[int]) – The size of feature maps.
  • gt_bboxes (Tensor) – Ground truth boxes, shape (n, 4).
  • stride (tuple(int)) – stride of current level
  • num_base_anchors (int) – The number of base anchors.
  • device (str, optional) – Device where the flags will be put on. Defaults to ‘cuda’.
Returns:

The valid flags of each anchor in a single level feature map.

Return type:

torch.Tensor

class mmdet.core.anchor.MlvlPointGenerator(strides, offset=0.5)[source]

Standard points generator for multi-level (Mlvl) feature maps in 2D points-based detectors.

Parameters:
  • strides (list[int] | list[tuple[int, int]]) – Strides of anchors in multiple feature levels in order (w, h).
  • offset (float) – The offset of points, the value is normalized with corresponding stride. Defaults to 0.5.
grid_priors(featmap_sizes, device='cuda', with_stride=False)[source]

Generate grid points of multiple feature levels.

Parameters:
  • featmap_sizes (list[tuple]) – List of feature map sizes in multiple feature levels, each size arrange as as (h, w).
  • device (str) – The device where the anchors will be put on.
  • with_stride (bool) – Whether to concatenate the stride to the last dimension of points.
Returns:

Points of multiple feature levels. The sizes of each tensor should be (N, 2) when with stride is False, where N = width * height, width and height are the sizes of the corresponding feature level, and the last dimension 2 represent (coord_x, coord_y), otherwise the shape should be (N, 4), and the last dimension 4 represent (coord_x, coord_y, stride_w, stride_h).

Return type:

list[torch.Tensor]

num_base_priors

The number of priors (points) at a point on the feature grid

Type:list[int]
num_levels

number of feature levels that the generator will be applied

Type:int
single_level_grid_priors(featmap_size, level_idx, device='cuda', with_stride=False)[source]

Generate grid Points of a single level.

Note

This function is usually called by method self.grid_priors.

Parameters:
  • featmap_size (tuple[int]) – Size of the feature maps, arrange as (h, w).
  • level_idx (int) – The index of corresponding feature map level.
  • device (str, optional) – The device the tensor will be put on. Defaults to ‘cuda’.
  • with_stride (bool) – Concatenate the stride to the last dimension of points.
Returns:

Points of single feature levels. The shape of tensor should be (N, 2) when with stride is False, where N = width * height, width and height are the sizes of the corresponding feature level, and the last dimension 2 represent (coord_x, coord_y), otherwise the shape should be (N, 4), and the last dimension 4 represent (coord_x, coord_y, stride_w, stride_h).

Return type:

Tensor

single_level_valid_flags(featmap_size, valid_size, device='cuda')[source]

Generate the valid flags of points of a single feature map.

Parameters:
  • featmap_size (tuple[int]) – The size of feature maps, arrange as as (h, w).
  • valid_size (tuple[int]) – The valid size of the feature maps. The size arrange as as (h, w).
  • device (str, optional) – The device where the flags will be put on. Defaults to ‘cuda’.
Returns:

The valid flags of each points in a single level feature map.

Return type:

torch.Tensor

sparse_priors(prior_idxs, featmap_size, level_idx, dtype=torch.float32, device='cuda')[source]

Generate sparse points according to the prior_idxs.

Parameters:
  • prior_idxs (Tensor) – The index of corresponding anchors in the feature map.
  • featmap_size (tuple[int]) – feature map size arrange as (w, h).
  • level_idx (int) – The level index of corresponding feature map.
  • (obj (device) – torch.dtype): Date type of points. Defaults to torch.float32.
  • (objtorch.device): The device where the points is located.
Returns:

Anchor with shape (N, 2), N should be equal to the length of prior_idxs. And last dimension 2 represent (coord_x, coord_y).

Return type:

Tensor

valid_flags(featmap_sizes, pad_shape, device='cuda')[source]

Generate valid flags of points of multiple feature levels.

Parameters:
  • featmap_sizes (list(tuple)) – List of feature map sizes in multiple feature levels, each size arrange as as (h, w).
  • pad_shape (tuple(int)) – The padded shape of the image, arrange as (h, w).
  • device (str) – The device where the anchors will be put on.
Returns:

Valid flags of points of multiple levels.

Return type:

list(torch.Tensor)

bbox

mmdet.core.bbox.bbox_overlaps(bboxes1, bboxes2, mode='iou', is_aligned=False, eps=1e-06)[source]

Calculate overlap between two set of bboxes.

FP16 Contributed by https://github.com/open-mmlab/mmdetection/pull/4889 .. note:

Assume bboxes1 is M x 4, bboxes2 is N x 4, when mode is 'iou',
there are some new generated variable when calculating IOU
using bbox_overlaps function:

1) is_aligned is False
    area1: M x 1
    area2: N x 1
    lt: M x N x 2
    rb: M x N x 2
    wh: M x N x 2
    overlap: M x N x 1
    union: M x N x 1
    ious: M x N x 1

    Total memory:
        S = (9 x N x M + N + M) * 4 Byte,

    When using FP16, we can reduce:
        R = (9 x N x M + N + M) * 4 / 2 Byte
        R large than (N + M) * 4 * 2 is always true when N and M >= 1.
        Obviously, N + M <= N * M < 3 * N * M, when N >=2 and M >=2,
                   N + 1 < 3 * N, when N or M is 1.

    Given M = 40 (ground truth), N = 400000 (three anchor boxes
    in per grid, FPN, R-CNNs),
        R = 275 MB (one times)

    A special case (dense detection), M = 512 (ground truth),
        R = 3516 MB = 3.43 GB

    When the batch size is B, reduce:
        B x R

    Therefore, CUDA memory runs out frequently.

    Experiments on GeForce RTX 2080Ti (11019 MiB):

    |   dtype   |   M   |   N   |   Use    |   Real   |   Ideal   |
    |:----:|:----:|:----:|:----:|:----:|:----:|
    |   FP32   |   512 | 400000 | 8020 MiB |   --   |   --   |
    |   FP16   |   512 | 400000 |   4504 MiB | 3516 MiB | 3516 MiB |
    |   FP32   |   40 | 400000 |   1540 MiB |   --   |   --   |
    |   FP16   |   40 | 400000 |   1264 MiB |   276MiB   | 275 MiB |

2) is_aligned is True
    area1: N x 1
    area2: N x 1
    lt: N x 2
    rb: N x 2
    wh: N x 2
    overlap: N x 1
    union: N x 1
    ious: N x 1

    Total memory:
        S = 11 x N * 4 Byte

    When using FP16, we can reduce:
        R = 11 x N * 4 / 2 Byte

So do the 'giou' (large than 'iou').

Time-wise, FP16 is generally faster than FP32.

When gpu_assign_thr is not -1, it takes more time on cpu
but not reduce memory.
There, we can reduce half the memory and keep the speed.

If is_aligned is False, then calculate the overlaps between each bbox of bboxes1 and bboxes2, otherwise the overlaps between each aligned pair of bboxes1 and bboxes2.

Parameters:
  • bboxes1 (Tensor) – shape (B, m, 4) in <x1, y1, x2, y2> format or empty.
  • bboxes2 (Tensor) – shape (B, n, 4) in <x1, y1, x2, y2> format or empty. B indicates the batch dim, in shape (B1, B2, …, Bn). If is_aligned is True, then m and n must be equal.
  • mode (str) – “iou” (intersection over union), “iof” (intersection over foreground) or “giou” (generalized intersection over union). Default “iou”.
  • is_aligned (bool, optional) – If True, then m and n must be equal. Default False.
  • eps (float, optional) – A value added to the denominator for numerical stability. Default 1e-6.
Returns:

shape (m, n) if is_aligned is False else shape (m,)

Return type:

Tensor

Example

>>> bboxes1 = torch.FloatTensor([
>>>     [0, 0, 10, 10],
>>>     [10, 10, 20, 20],
>>>     [32, 32, 38, 42],
>>> ])
>>> bboxes2 = torch.FloatTensor([
>>>     [0, 0, 10, 20],
>>>     [0, 10, 10, 19],
>>>     [10, 10, 20, 20],
>>> ])
>>> overlaps = bbox_overlaps(bboxes1, bboxes2)
>>> assert overlaps.shape == (3, 3)
>>> overlaps = bbox_overlaps(bboxes1, bboxes2, is_aligned=True)
>>> assert overlaps.shape == (3, )

Example

>>> empty = torch.empty(0, 4)
>>> nonempty = torch.FloatTensor([[0, 0, 10, 9]])
>>> assert tuple(bbox_overlaps(empty, nonempty).shape) == (0, 1)
>>> assert tuple(bbox_overlaps(nonempty, empty).shape) == (1, 0)
>>> assert tuple(bbox_overlaps(empty, empty).shape) == (0, 0)
class mmdet.core.bbox.BboxOverlaps2D(scale=1.0, dtype=None)[source]

2D Overlaps (e.g. IoUs, GIoUs) Calculator.

class mmdet.core.bbox.BaseAssigner[source]

Base assigner that assigns boxes to ground truth boxes.

assign(bboxes, gt_bboxes, gt_bboxes_ignore=None, gt_labels=None)[source]

Assign boxes to either a ground truth boxes or a negative boxes.

class mmdet.core.bbox.MaxIoUAssigner(pos_iou_thr, neg_iou_thr, min_pos_iou=0.0, gt_max_assign_all=True, ignore_iof_thr=-1, ignore_wrt_candidates=True, match_low_quality=True, gpu_assign_thr=-1, iou_calculator={'type': 'BboxOverlaps2D'})[source]

Assign a corresponding gt bbox or background to each bbox.

Each proposals will be assigned with -1, or a semi-positive integer indicating the ground truth index.

  • -1: negative sample, no assigned gt
  • semi-positive integer: positive sample, index (0-based) of assigned gt
Parameters:
  • pos_iou_thr (float) – IoU threshold for positive bboxes.
  • neg_iou_thr (float or tuple) – IoU threshold for negative bboxes.
  • min_pos_iou (float) – Minimum iou for a bbox to be considered as a positive bbox. Positive samples can have smaller IoU than pos_iou_thr due to the 4th step (assign max IoU sample to each gt).
  • gt_max_assign_all (bool) – Whether to assign all bboxes with the same highest overlap with some gt to that gt.
  • ignore_iof_thr (float) – IoF threshold for ignoring bboxes (if gt_bboxes_ignore is specified). Negative values mean not ignoring any bboxes.
  • ignore_wrt_candidates (bool) – Whether to compute the iof between bboxes and gt_bboxes_ignore, or the contrary.
  • match_low_quality (bool) – Whether to allow low quality matches. This is usually allowed for RPN and single stage detectors, but not allowed in the second stage. Details are demonstrated in Step 4.
  • gpu_assign_thr (int) – The upper bound of the number of GT for GPU assign. When the number of gt is above this threshold, will assign on CPU device. Negative values mean not assign on CPU.
assign(bboxes, gt_bboxes, gt_bboxes_ignore=None, gt_labels=None)[source]

Assign gt to bboxes.

This method assign a gt bbox to every bbox (proposal/anchor), each bbox will be assigned with -1, or a semi-positive number. -1 means negative sample, semi-positive number is the index (0-based) of assigned gt. The assignment is done in following steps, the order matters.

  1. assign every bbox to the background
  2. assign proposals whose iou with all gts < neg_iou_thr to 0
  3. for each bbox, if the iou with its nearest gt >= pos_iou_thr, assign it to that bbox
  4. for each gt bbox, assign its nearest proposals (may be more than one) to itself
Parameters:
  • bboxes (Tensor) – Bounding boxes to be assigned, shape(n, 4).
  • gt_bboxes (Tensor) – Groundtruth boxes, shape (k, 4).
  • gt_bboxes_ignore (Tensor, optional) – Ground truth bboxes that are labelled as ignored, e.g., crowd boxes in COCO.
  • gt_labels (Tensor, optional) – Label of gt_bboxes, shape (k, ).
Returns:

The assign result.

Return type:

AssignResult

Example

>>> self = MaxIoUAssigner(0.5, 0.5)
>>> bboxes = torch.Tensor([[0, 0, 10, 10], [10, 10, 20, 20]])
>>> gt_bboxes = torch.Tensor([[0, 0, 10, 9]])
>>> assign_result = self.assign(bboxes, gt_bboxes)
>>> expected_gt_inds = torch.LongTensor([1, 0])
>>> assert torch.all(assign_result.gt_inds == expected_gt_inds)
assign_wrt_overlaps(overlaps, gt_labels=None)[source]

Assign w.r.t. the overlaps of bboxes with gts.

Parameters:
  • overlaps (Tensor) – Overlaps between k gt_bboxes and n bboxes, shape(k, n).
  • gt_labels (Tensor, optional) – Labels of k gt_bboxes, shape (k, ).
Returns:

The assign result.

Return type:

AssignResult

class mmdet.core.bbox.AssignResult(num_gts, gt_inds, max_overlaps, labels=None)[source]

Stores assignments between predicted and truth boxes.

num_gts

the number of truth boxes considered when computing this assignment

Type:int
gt_inds

for each predicted box indicates the 1-based index of the assigned truth box. 0 means unassigned and -1 means ignore.

Type:LongTensor
max_overlaps

the iou between the predicted box and its assigned truth box.

Type:FloatTensor
labels

If specified, for each predicted box indicates the category label of the assigned truth box.

Type:None | LongTensor

Example

>>> # An assign result between 4 predicted boxes and 9 true boxes
>>> # where only two boxes were assigned.
>>> num_gts = 9
>>> max_overlaps = torch.LongTensor([0, .5, .9, 0])
>>> gt_inds = torch.LongTensor([-1, 1, 2, 0])
>>> labels = torch.LongTensor([0, 3, 4, 0])
>>> self = AssignResult(num_gts, gt_inds, max_overlaps, labels)
>>> print(str(self))  # xdoctest: +IGNORE_WANT
<AssignResult(num_gts=9, gt_inds.shape=(4,), max_overlaps.shape=(4,),
              labels.shape=(4,))>
>>> # Force addition of gt labels (when adding gt as proposals)
>>> new_labels = torch.LongTensor([3, 4, 5])
>>> self.add_gt_(new_labels)
>>> print(str(self))  # xdoctest: +IGNORE_WANT
<AssignResult(num_gts=9, gt_inds.shape=(7,), max_overlaps.shape=(7,),
              labels.shape=(7,))>
add_gt_(gt_labels)[source]

Add ground truth as assigned results.

Parameters:gt_labels (torch.Tensor) – Labels of gt boxes
get_extra_property(key)[source]

Get user-defined property.

info

a dictionary of info about the object

Type:dict
num_preds

the number of predictions in this assignment

Type:int
classmethod random(**kwargs)[source]

Create random AssignResult for tests or debugging.

Parameters:
  • num_preds – number of predicted boxes
  • num_gts – number of true boxes
  • p_ignore (float) – probability of a predicted box assigned to an ignored truth
  • p_assigned (float) – probability of a predicted box not being assigned
  • p_use_label (float | bool) – with labels or not
  • rng (None | int | numpy.random.RandomState) – seed or state
Returns:

Randomly generated assign results.

Return type:

AssignResult

Example

>>> from mmdet.core.bbox.assigners.assign_result import *  # NOQA
>>> self = AssignResult.random()
>>> print(self.info)
set_extra_property(key, value)[source]

Set user-defined new property.

class mmdet.core.bbox.BaseSampler(num, pos_fraction, neg_pos_ub=-1, add_gt_as_proposals=True, **kwargs)[source]

Base class of samplers.

sample(assign_result, bboxes, gt_bboxes, gt_labels=None, **kwargs)[source]

Sample positive and negative bboxes.

This is a simple implementation of bbox sampling given candidates, assigning results and ground truth bboxes.

Parameters:
  • assign_result (AssignResult) – Bbox assigning results.
  • bboxes (Tensor) – Boxes to be sampled from.
  • gt_bboxes (Tensor) – Ground truth bboxes.
  • gt_labels (Tensor, optional) – Class labels of ground truth bboxes.
Returns:

Sampling result.

Return type:

SamplingResult

Example

>>> from mmdet.core.bbox import RandomSampler
>>> from mmdet.core.bbox import AssignResult
>>> from mmdet.core.bbox.demodata import ensure_rng, random_boxes
>>> rng = ensure_rng(None)
>>> assign_result = AssignResult.random(rng=rng)
>>> bboxes = random_boxes(assign_result.num_preds, rng=rng)
>>> gt_bboxes = random_boxes(assign_result.num_gts, rng=rng)
>>> gt_labels = None
>>> self = RandomSampler(num=32, pos_fraction=0.5, neg_pos_ub=-1,
>>>                      add_gt_as_proposals=False)
>>> self = self.sample(assign_result, bboxes, gt_bboxes, gt_labels)
class mmdet.core.bbox.PseudoSampler(**kwargs)[source]

A pseudo sampler that does not do sampling actually.

sample(assign_result, bboxes, gt_bboxes, **kwargs)[source]

Directly returns the positive and negative indices of samples.

Parameters:
  • assign_result (AssignResult) – Assigned results
  • bboxes (torch.Tensor) – Bounding boxes
  • gt_bboxes (torch.Tensor) – Ground truth boxes
Returns:

sampler results

Return type:

SamplingResult

class mmdet.core.bbox.RandomSampler(num, pos_fraction, neg_pos_ub=-1, add_gt_as_proposals=True, **kwargs)[source]

Random sampler.

Parameters:
  • num (int) – Number of samples
  • pos_fraction (float) – Fraction of positive samples
  • neg_pos_up (int, optional) – Upper bound number of negative and positive samples. Defaults to -1.
  • add_gt_as_proposals (bool, optional) – Whether to add ground truth boxes as proposals. Defaults to True.
random_choice(gallery, num)[source]

Random select some elements from the gallery.

If gallery is a Tensor, the returned indices will be a Tensor; If gallery is a ndarray or list, the returned indices will be a ndarray.

Parameters:
  • gallery (Tensor | ndarray | list) – indices pool.
  • num (int) – expected sample num.
Returns:

sampled indices.

Return type:

Tensor or ndarray

class mmdet.core.bbox.InstanceBalancedPosSampler(num, pos_fraction, neg_pos_ub=-1, add_gt_as_proposals=True, **kwargs)[source]

Instance balanced sampler that samples equal number of positive samples for each instance.

class mmdet.core.bbox.IoUBalancedNegSampler(num, pos_fraction, floor_thr=-1, floor_fraction=0, num_bins=3, **kwargs)[source]

IoU Balanced Sampling.

arXiv: https://arxiv.org/pdf/1904.02701.pdf (CVPR 2019)

Sampling proposals according to their IoU. floor_fraction of needed RoIs are sampled from proposals whose IoU are lower than floor_thr randomly. The others are sampled from proposals whose IoU are higher than floor_thr. These proposals are sampled from some bins evenly, which are split by num_bins via IoU evenly.

Parameters:
  • num (int) – number of proposals.
  • pos_fraction (float) – fraction of positive proposals.
  • floor_thr (float) – threshold (minimum) IoU for IoU balanced sampling, set to -1 if all using IoU balanced sampling.
  • floor_fraction (float) – sampling fraction of proposals under floor_thr.
  • num_bins (int) – number of bins in IoU balanced sampling.
sample_via_interval(max_overlaps, full_set, num_expected)[source]

Sample according to the iou interval.

Parameters:
  • max_overlaps (torch.Tensor) – IoU between bounding boxes and ground truth boxes.
  • full_set (set(int)) – A full set of indices of boxes。
  • num_expected (int) – Number of expected samples。
Returns:

Indices of samples

Return type:

np.ndarray

class mmdet.core.bbox.CombinedSampler(pos_sampler, neg_sampler, **kwargs)[source]

A sampler that combines positive sampler and negative sampler.

class mmdet.core.bbox.OHEMSampler(num, pos_fraction, context, neg_pos_ub=-1, add_gt_as_proposals=True, **kwargs)[source]

Online Hard Example Mining Sampler described in Training Region-based Object Detectors with Online Hard Example Mining.

class mmdet.core.bbox.SamplingResult(pos_inds, neg_inds, bboxes, gt_bboxes, assign_result, gt_flags)[source]

Bbox sampling result.

Example

>>> # xdoctest: +IGNORE_WANT
>>> from mmdet.core.bbox.samplers.sampling_result import *  # NOQA
>>> self = SamplingResult.random(rng=10)
>>> print(f'self = {self}')
self = <SamplingResult({
    'neg_bboxes': torch.Size([12, 4]),
    'neg_inds': tensor([ 0,  1,  2,  4,  5,  6,  7,  8,  9, 10, 11, 12]),
    'num_gts': 4,
    'pos_assigned_gt_inds': tensor([], dtype=torch.int64),
    'pos_bboxes': torch.Size([0, 4]),
    'pos_inds': tensor([], dtype=torch.int64),
    'pos_is_gt': tensor([], dtype=torch.uint8)
})>
bboxes

concatenated positive and negative boxes

Type:torch.Tensor
info

Returns a dictionary of info about the object.

classmethod random(rng=None, **kwargs)[source]
Parameters:
  • rng (None | int | numpy.random.RandomState) – seed or state.
  • kwargs (keyword arguments) –
    • num_preds: number of predicted boxes
    • num_gts: number of true boxes
    • p_ignore (float): probability of a predicted box assigned to an ignored truth.
    • p_assigned (float): probability of a predicted box not being assigned.
    • p_use_label (float | bool): with labels or not.
Returns:

Randomly generated sampling result.

Return type:

SamplingResult

Example

>>> from mmdet.core.bbox.samplers.sampling_result import *  # NOQA
>>> self = SamplingResult.random()
>>> print(self.__dict__)
to(device)[source]

Change the device of the data inplace.

Example

>>> self = SamplingResult.random()
>>> print(f'self = {self.to(None)}')
>>> # xdoctest: +REQUIRES(--gpu)
>>> print(f'self = {self.to(0)}')
class mmdet.core.bbox.ScoreHLRSampler(num, pos_fraction, context, neg_pos_ub=-1, add_gt_as_proposals=True, k=0.5, bias=0, score_thr=0.05, iou_thr=0.5, **kwargs)[source]

Importance-based Sample Reweighting (ISR_N), described in Prime Sample Attention in Object Detection.

Score hierarchical local rank (HLR) differentiates with RandomSampler in negative part. It firstly computes Score-HLR in a two-step way, then linearly maps score hlr to the loss weights.

Parameters:
  • num (int) – Total number of sampled RoIs.
  • pos_fraction (float) – Fraction of positive samples.
  • context (BaseRoIHead) – RoI head that the sampler belongs to.
  • neg_pos_ub (int) – Upper bound of the ratio of num negative to num positive, -1 means no upper bound.
  • add_gt_as_proposals (bool) – Whether to add ground truth as proposals.
  • k (float) – Power of the non-linear mapping.
  • bias (float) – Shift of the non-linear mapping.
  • score_thr (float) – Minimum score that a negative sample is to be considered as valid bbox.
static random_choice(gallery, num)[source]

Randomly select some elements from the gallery.

If gallery is a Tensor, the returned indices will be a Tensor; If gallery is a ndarray or list, the returned indices will be a ndarray.

Parameters:
  • gallery (Tensor | ndarray | list) – indices pool.
  • num (int) – expected sample num.
Returns:

sampled indices.

Return type:

Tensor or ndarray

sample(assign_result, bboxes, gt_bboxes, gt_labels=None, img_meta=None, **kwargs)[source]

Sample positive and negative bboxes.

This is a simple implementation of bbox sampling given candidates, assigning results and ground truth bboxes.

Parameters:
  • assign_result (AssignResult) – Bbox assigning results.
  • bboxes (Tensor) – Boxes to be sampled from.
  • gt_bboxes (Tensor) – Ground truth bboxes.
  • gt_labels (Tensor, optional) – Class labels of ground truth bboxes.
Returns:

Sampling result and negative

label weights.

Return type:

tuple[SamplingResult, Tensor]

mmdet.core.bbox.build_assigner(cfg, **default_args)[source]

Builder of box assigner.

mmdet.core.bbox.build_sampler(cfg, **default_args)[source]

Builder of box sampler.

mmdet.core.bbox.bbox_flip(bboxes, img_shape, direction='horizontal')[source]

Flip bboxes horizontally or vertically.

Parameters:
  • bboxes (Tensor) – Shape (…, 4*k)
  • img_shape (tuple) – Image shape.
  • direction (str) – Flip direction, options are “horizontal”, “vertical”, “diagonal”. Default: “horizontal”
Returns:

Flipped bboxes.

Return type:

Tensor

mmdet.core.bbox.bbox_mapping(bboxes, img_shape, scale_factor, flip, flip_direction='horizontal')[source]

Map bboxes from the original image scale to testing scale.

mmdet.core.bbox.bbox_mapping_back(bboxes, img_shape, scale_factor, flip, flip_direction='horizontal')[source]

Map bboxes from testing scale to original image scale.

mmdet.core.bbox.bbox2roi(bbox_list)[source]

Convert a list of bboxes to roi format.

Parameters:bbox_list (list[Tensor]) – a list of bboxes corresponding to a batch of images.
Returns:shape (n, 5), [batch_ind, x1, y1, x2, y2]
Return type:Tensor
mmdet.core.bbox.roi2bbox(rois)[source]

Convert rois to bounding box format.

Parameters:rois (torch.Tensor) – RoIs with the shape (n, 5) where the first column indicates batch id of each RoI.
Returns:Converted boxes of corresponding rois.
Return type:list[torch.Tensor]
mmdet.core.bbox.bbox2result(bboxes, labels, num_classes)[source]

Convert detection results to a list of numpy arrays.

Parameters:
  • bboxes (torch.Tensor | np.ndarray) – shape (n, 5)
  • labels (torch.Tensor | np.ndarray) – shape (n, )
  • num_classes (int) – class number, including background class
Returns:

bbox results of each class

Return type:

list(ndarray)

mmdet.core.bbox.distance2bbox(points, distance, max_shape=None)[source]

Decode distance prediction to bounding box.

Parameters:
  • points (Tensor) – Shape (B, N, 2) or (N, 2).
  • distance (Tensor) – Distance from the given point to 4 boundaries (left, top, right, bottom). Shape (B, N, 4) or (N, 4)
  • (Sequence[int] or torch.Tensor or Sequence[ (max_shape) – Sequence[int]],optional): Maximum bounds for boxes, specifies (H, W, C) or (H, W). If priors shape is (B, N, 4), then the max_shape should be a Sequence[Sequence[int]] and the length of max_shape should also be B.
Returns:

Boxes with shape (N, 4) or (B, N, 4)

Return type:

Tensor

mmdet.core.bbox.bbox2distance(points, bbox, max_dis=None, eps=0.1)[source]

Decode bounding box based on distances.

Parameters:
  • points (Tensor) – Shape (n, 2), [x, y].
  • bbox (Tensor) – Shape (n, 4), “xyxy” format
  • max_dis (float) – Upper bound of the distance.
  • eps (float) – a small value to ensure target < max_dis, instead <=
Returns:

Decoded distances.

Return type:

Tensor

mmdet.core.bbox.build_bbox_coder(cfg, **default_args)[source]

Builder of box coder.

class mmdet.core.bbox.BaseBBoxCoder(**kwargs)[source]

Base bounding box coder.

decode(bboxes, bboxes_pred)[source]

Decode the predicted bboxes according to prediction and base boxes.

encode(bboxes, gt_bboxes)[source]

Encode deltas between bboxes and ground truth boxes.

class mmdet.core.bbox.PseudoBBoxCoder(**kwargs)[source]

Pseudo bounding box coder.

decode(bboxes, pred_bboxes)[source]

torch.Tensor: return the given pred_bboxes

encode(bboxes, gt_bboxes)[source]

torch.Tensor: return the given bboxes

class mmdet.core.bbox.DeltaXYWHBBoxCoder(target_means=(0.0, 0.0, 0.0, 0.0), target_stds=(1.0, 1.0, 1.0, 1.0), clip_border=True, add_ctr_clamp=False, ctr_clamp=32)[source]

Delta XYWH BBox coder.

Following the practice in R-CNN, this coder encodes bbox (x1, y1, x2, y2) into delta (dx, dy, dw, dh) and decodes delta (dx, dy, dw, dh) back to original bbox (x1, y1, x2, y2).

Parameters:
  • target_means (Sequence[float]) – Denormalizing means of target for delta coordinates
  • target_stds (Sequence[float]) – Denormalizing standard deviation of target for delta coordinates
  • clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.
  • add_ctr_clamp (bool) – Whether to add center clamp, when added, the predicted box is clamped is its center is too far away from the original anchor’s center. Only used by YOLOF. Default False.
  • ctr_clamp (int) – the maximum pixel shift to clamp. Only used by YOLOF. Default 32.
decode(bboxes, pred_bboxes, max_shape=None, wh_ratio_clip=0.016)[source]

Apply transformation pred_bboxes to boxes.

Parameters:
  • bboxes (torch.Tensor) – Basic boxes. Shape (B, N, 4) or (N, 4)
  • pred_bboxes (Tensor) – Encoded offsets with respect to each roi. Has shape (B, N, num_classes * 4) or (B, N, 4) or (N, num_classes * 4) or (N, 4). Note N = num_anchors * W * H when rois is a grid of anchors.Offset encoding follows [1].
  • (Sequence[int] or torch.Tensor or Sequence[ (max_shape) – Sequence[int]],optional): Maximum bounds for boxes, specifies (H, W, C) or (H, W). If bboxes shape is (B, N, 4), then the max_shape should be a Sequence[Sequence[int]] and the length of max_shape should also be B.
  • wh_ratio_clip (float, optional) – The allowed ratio between width and height.
Returns:

Decoded boxes.

Return type:

torch.Tensor

encode(bboxes, gt_bboxes)[source]

Get box regression transformation deltas that can be used to transform the bboxes into the gt_bboxes.

Parameters:
  • bboxes (torch.Tensor) – Source boxes, e.g., object proposals.
  • gt_bboxes (torch.Tensor) – Target of the transformation, e.g., ground-truth boxes.
Returns:

Box transformation deltas

Return type:

torch.Tensor

class mmdet.core.bbox.TBLRBBoxCoder(normalizer=4.0, clip_border=True)[source]

TBLR BBox coder.

Following the practice in FSAF, this coder encodes gt bboxes (x1, y1, x2, y2) into (top, bottom, left, right) and decode it back to the original.

Parameters:
  • normalizer (list | float) – Normalization factor to be divided with when coding the coordinates. If it is a list, it should have length of 4 indicating normalization factor in tblr dims. Otherwise it is a unified float factor for all dims. Default: 4.0
  • clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.
decode(bboxes, pred_bboxes, max_shape=None)[source]

Apply transformation pred_bboxes to boxes.

Parameters:
  • bboxes (torch.Tensor) – Basic boxes.Shape (B, N, 4) or (N, 4)
  • pred_bboxes (torch.Tensor) – Encoded boxes with shape (B, N, 4) or (N, 4)
  • (Sequence[int] or torch.Tensor or Sequence[ (max_shape) – Sequence[int]],optional): Maximum bounds for boxes, specifies (H, W, C) or (H, W). If bboxes shape is (B, N, 4), then the max_shape should be a Sequence[Sequence[int]] and the length of max_shape should also be B.
Returns:

Decoded boxes.

Return type:

torch.Tensor

encode(bboxes, gt_bboxes)[source]

Get box regression transformation deltas that can be used to transform the bboxes into the gt_bboxes in the (top, left, bottom, right) order.

Parameters:
  • bboxes (torch.Tensor) – source boxes, e.g., object proposals.
  • gt_bboxes (torch.Tensor) – target of the transformation, e.g., ground truth boxes.
Returns:

Box transformation deltas

Return type:

torch.Tensor

class mmdet.core.bbox.CenterRegionAssigner(pos_scale, neg_scale, min_pos_iof=0.01, ignore_gt_scale=0.5, foreground_dominate=False, iou_calculator={'type': 'BboxOverlaps2D'})[source]

Assign pixels at the center region of a bbox as positive.

Each proposals will be assigned with -1, 0, or a positive integer indicating the ground truth index. - -1: negative samples - semi-positive numbers: positive sample, index (0-based) of assigned gt

Parameters:
  • pos_scale (float) – Threshold within which pixels are labelled as positive.
  • neg_scale (float) – Threshold above which pixels are labelled as positive.
  • min_pos_iof (float) – Minimum iof of a pixel with a gt to be labelled as positive. Default: 1e-2
  • ignore_gt_scale (float) – Threshold within which the pixels are ignored when the gt is labelled as shadowed. Default: 0.5
  • foreground_dominate (bool) – If True, the bbox will be assigned as positive when a gt’s kernel region overlaps with another’s shadowed (ignored) region, otherwise it is set as ignored. Default to False.
assign(bboxes, gt_bboxes, gt_bboxes_ignore=None, gt_labels=None)[source]

Assign gt to bboxes.

This method assigns gts to every bbox (proposal/anchor), each bbox will be assigned with -1, or a semi-positive number. -1 means negative sample, semi-positive number is the index (0-based) of assigned gt.

Parameters:
  • bboxes (Tensor) – Bounding boxes to be assigned, shape(n, 4).
  • gt_bboxes (Tensor) – Groundtruth boxes, shape (k, 4).
  • gt_bboxes_ignore (tensor, optional) – Ground truth bboxes that are labelled as ignored, e.g., crowd boxes in COCO.
  • gt_labels (tensor, optional) – Label of gt_bboxes, shape (num_gts,).
Returns:

The assigned result. Note that shadowed_labels of shape (N, 2) is also added as an assign_result attribute. shadowed_labels is a tensor composed of N pairs of anchor_ind, class_label], where N is the number of anchors that lie in the outer region of a gt, anchor_ind is the shadowed anchor index and class_label is the shadowed class label.

Return type:

AssignResult

Example

>>> self = CenterRegionAssigner(0.2, 0.2)
>>> bboxes = torch.Tensor([[0, 0, 10, 10], [10, 10, 20, 20]])
>>> gt_bboxes = torch.Tensor([[0, 0, 10, 10]])
>>> assign_result = self.assign(bboxes, gt_bboxes)
>>> expected_gt_inds = torch.LongTensor([1, 0])
>>> assert torch.all(assign_result.gt_inds == expected_gt_inds)
assign_one_hot_gt_indices(is_bbox_in_gt_core, is_bbox_in_gt_shadow, gt_priority=None)[source]

Assign only one gt index to each prior box.

Gts with large gt_priority are more likely to be assigned.

Parameters:
  • is_bbox_in_gt_core (Tensor) – Bool tensor indicating the bbox center is in the core area of a gt (e.g. 0-0.2). Shape: (num_prior, num_gt).
  • is_bbox_in_gt_shadow (Tensor) – Bool tensor indicating the bbox center is in the shadowed area of a gt (e.g. 0.2-0.5). Shape: (num_prior, num_gt).
  • gt_priority (Tensor) – Priorities of gts. The gt with a higher priority is more likely to be assigned to the bbox when the bbox match with multiple gts. Shape: (num_gt, ).
Returns:

Returns (assigned_gt_inds, shadowed_gt_inds).

  • assigned_gt_inds: The assigned gt index of each prior bbox (i.e. index from 1 to num_gts). Shape: (num_prior, ).
  • shadowed_gt_inds: shadowed gt indices. It is a tensor of shape (num_ignore, 2) with first column being the shadowed prior bbox indices and the second column the shadowed gt indices (1-based).

Return type:

tuple

get_gt_priorities(gt_bboxes)[source]

Get gt priorities according to their areas.

Smaller gt has higher priority.

Parameters:gt_bboxes (Tensor) – Ground truth boxes, shape (k, 4).
Returns:The priority of gts so that gts with larger priority is more likely to be assigned. Shape (k, )
Return type:Tensor
mmdet.core.bbox.bbox_rescale(bboxes, scale_factor=1.0)[source]

Rescale bounding box w.r.t. scale_factor.

Parameters:
  • bboxes (Tensor) – Shape (n, 4) for bboxes or (n, 5) for rois
  • scale_factor (float) – rescale factor
Returns:

Rescaled bboxes.

Return type:

Tensor

mmdet.core.bbox.bbox_cxcywh_to_xyxy(bbox)[source]

Convert bbox coordinates from (cx, cy, w, h) to (x1, y1, x2, y2).

Parameters:bbox (Tensor) – Shape (n, 4) for bboxes.
Returns:Converted bboxes.
Return type:Tensor
mmdet.core.bbox.bbox_xyxy_to_cxcywh(bbox)[source]

Convert bbox coordinates from (x1, y1, x2, y2) to (cx, cy, w, h).

Parameters:bbox (Tensor) – Shape (n, 4) for bboxes.
Returns:Converted bboxes.
Return type:Tensor
class mmdet.core.bbox.RegionAssigner(center_ratio=0.2, ignore_ratio=0.5)[source]

Assign a corresponding gt bbox or background to each bbox.

Each proposals will be assigned with -1, 0, or a positive integer indicating the ground truth index.

  • -1: don’t care
  • 0: negative sample, no assigned gt
  • positive integer: positive sample, index (1-based) of assigned gt
Parameters:
  • center_ratio – ratio of the region in the center of the bbox to define positive sample.
  • ignore_ratio – ratio of the region to define ignore samples.
assign(mlvl_anchors, mlvl_valid_flags, gt_bboxes, img_meta, featmap_sizes, anchor_scale, anchor_strides, gt_bboxes_ignore=None, gt_labels=None, allowed_border=0)[source]

Assign gt to anchors.

This method assign a gt bbox to every bbox (proposal/anchor), each bbox will be assigned with -1, 0, or a positive number. -1 means don’t care, 0 means negative sample, positive number is the index (1-based) of assigned gt.

The assignment is done in following steps, and the order matters.

  1. Assign every anchor to 0 (negative)
  2. (For each gt_bboxes) Compute ignore flags based on ignore_region then assign -1 to anchors w.r.t. ignore flags
  3. (For each gt_bboxes) Compute pos flags based on center_region then assign gt_bboxes to anchors w.r.t. pos flags
  4. (For each gt_bboxes) Compute ignore flags based on adjacent anchor level then assign -1 to anchors w.r.t. ignore flags
  5. Assign anchor outside of image to -1
Parameters:
  • mlvl_anchors (list[Tensor]) – Multi level anchors.
  • mlvl_valid_flags (list[Tensor]) – Multi level valid flags.
  • gt_bboxes (Tensor) – Ground truth bboxes of image
  • img_meta (dict) – Meta info of image.
  • featmap_sizes (list[Tensor]) – Feature mapsize each level
  • anchor_scale (int) – Scale of the anchor.
  • anchor_strides (list[int]) – Stride of the anchor.
  • gt_bboxes – Groundtruth boxes, shape (k, 4).
  • gt_bboxes_ignore (Tensor, optional) – Ground truth bboxes that are labelled as ignored, e.g., crowd boxes in COCO.
  • gt_labels (Tensor, optional) – Label of gt_bboxes, shape (k, ).
  • allowed_border (int, optional) – The border to allow the valid anchor. Defaults to 0.
Returns:

The assign result.

Return type:

AssignResult

export

mmdet.core.export.build_model_from_cfg(config_path, checkpoint_path, cfg_options=None)[source]

Build a model from config and load the given checkpoint.

Parameters:
  • config_path (str) – the OpenMMLab config for the model we want to export to ONNX
  • checkpoint_path (str) – Path to the corresponding checkpoint
Returns:

the built model

Return type:

torch.nn.Module

mmdet.core.export.generate_inputs_and_wrap_model(config_path, checkpoint_path, input_config, cfg_options=None)[source]

Prepare sample input and wrap model for ONNX export.

The ONNX export API only accept args, and all inputs should be torch.Tensor or corresponding types (such as tuple of tensor). So we should call this function before exporting. This function will:

  1. generate corresponding inputs which are used to execute the model.
  2. Wrap the model’s forward function.

For example, the MMDet models’ forward function has a parameter return_loss:bool. As we want to set it as False while export API supports neither bool type or kwargs. So we have to replace the forward method like model.forward = partial(model.forward, return_loss=False).

Parameters:
  • config_path (str) – the OpenMMLab config for the model we want to export to ONNX
  • checkpoint_path (str) – Path to the corresponding checkpoint
  • input_config (dict) – the exactly data in this dict depends on the framework. For MMSeg, we can just declare the input shape, and generate the dummy data accordingly. However, for MMDet, we may pass the real img path, or the NMS will return None as there is no legal bbox.
Returns:

(model, tensor_data) wrapped model which can be called by

model(*tensor_data) and a list of inputs which are used to execute the model while exporting.

Return type:

tuple

mmdet.core.export.preprocess_example_input(input_config)[source]

Prepare an example input image for generate_inputs_and_wrap_model.

Parameters:input_config (dict) – customized config describing the example input.
Returns:(one_img, one_meta), tensor of the example input image and meta information for the example input image.
Return type:tuple

Examples

>>> from mmdet.core.export import preprocess_example_input
>>> input_config = {
>>>         'input_shape': (1,3,224,224),
>>>         'input_path': 'demo/demo.jpg',
>>>         'normalize_cfg': {
>>>             'mean': (123.675, 116.28, 103.53),
>>>             'std': (58.395, 57.12, 57.375)
>>>             }
>>>         }
>>> one_img, one_meta = preprocess_example_input(input_config)
>>> print(one_img.shape)
torch.Size([1, 3, 224, 224])
>>> print(one_meta)
{'img_shape': (224, 224, 3),
'ori_shape': (224, 224, 3),
'pad_shape': (224, 224, 3),
'filename': '<demo>.png',
'scale_factor': 1.0,
'flip': False}
mmdet.core.export.get_k_for_topk(k, size)[source]

Get k of TopK for onnx exporting.

The K of TopK in TensorRT should not be a Tensor, while in ONNX Runtime
it could be a Tensor.Due to dynamic shape feature, we have to decide whether to do TopK and what K it should be while exporting to ONNX.
If returned K is less than zero, it means we do not have to do
TopK operation.
Parameters:
  • k (int or Tensor) – The set k value for nms from config file.
  • size (Tensor or torch.Size) – The number of elements of TopK’s input tensor
Returns:

(int or Tensor): The final K for TopK.

Return type:

tuple

mmdet.core.export.add_dummy_nms_for_onnx(boxes, scores, max_output_boxes_per_class=1000, iou_threshold=0.5, score_threshold=0.05, pre_top_k=-1, after_top_k=-1, labels=None)[source]

Create a dummy onnx::NonMaxSuppression op while exporting to ONNX.

This function helps exporting to onnx with batch and multiclass NMS op. It only supports class-agnostic detection results. That is, the scores is of shape (N, num_bboxes, num_classes) and the boxes is of shape (N, num_boxes, 4).

Parameters:
  • boxes (Tensor) – The bounding boxes of shape [N, num_boxes, 4]
  • scores (Tensor) – The detection scores of shape [N, num_boxes, num_classes]
  • max_output_boxes_per_class (int) – Maximum number of output boxes per class of nms. Defaults to 1000.
  • iou_threshold (float) – IOU threshold of nms. Defaults to 0.5
  • score_threshold (float) – score threshold of nms. Defaults to 0.05.
  • pre_top_k (bool) – Number of top K boxes to keep before nms. Defaults to -1.
  • after_top_k (int) – Number of top K boxes to keep after nms. Defaults to -1.
  • labels (Tensor, optional) – It not None, explicit labels would be used. Otherwise, labels would be automatically generated using num_classed. Defaults to None.
Returns:

dets of shape [N, num_det, 5] and class labels

of shape [N, num_det].

Return type:

tuple[Tensor, Tensor]

mmdet.core.export.dynamic_clip_for_onnx(x1, y1, x2, y2, max_shape)[source]

Clip boxes dynamically for onnx.

Since torch.clamp cannot have dynamic min and max, we scale the
boxes by 1/max_shape and clamp in the range [0, 1].
Parameters:
  • x1 (Tensor) – The x1 for bounding boxes.
  • y1 (Tensor) – The y1 for bounding boxes.
  • x2 (Tensor) – The x2 for bounding boxes.
  • y2 (Tensor) – The y2 for bounding boxes.
  • max_shape (Tensor or torch.Size) – The (H,W) of original image.
Returns:

The clipped x1, y1, x2, y2.

Return type:

tuple(Tensor)

mask

mmdet.core.mask.split_combined_polys(polys, poly_lens, polys_per_mask)[source]

Split the combined 1-D polys into masks.

A mask is represented as a list of polys, and a poly is represented as a 1-D array. In dataset, all masks are concatenated into a single 1-D tensor. Here we need to split the tensor into original representations.

Parameters:
  • polys (list) – a list (length = image num) of 1-D tensors
  • poly_lens (list) – a list (length = image num) of poly length
  • polys_per_mask (list) – a list (length = image num) of poly number of each mask
Returns:

a list (length = image num) of list (length = mask num) of list (length = poly num) of numpy array.

Return type:

list

mmdet.core.mask.mask_target(pos_proposals_list, pos_assigned_gt_inds_list, gt_masks_list, cfg)[source]

Compute mask target for positive proposals in multiple images.

Parameters:
  • pos_proposals_list (list[Tensor]) – Positive proposals in multiple images.
  • pos_assigned_gt_inds_list (list[Tensor]) – Assigned GT indices for each positive proposals.
  • gt_masks_list (list[BaseInstanceMasks]) – Ground truth masks of each image.
  • cfg (dict) – Config dict that specifies the mask size.
Returns:

Mask target of each image.

Return type:

list[Tensor]

Example

>>> import mmcv
>>> import mmdet
>>> from mmdet.core.mask import BitmapMasks
>>> from mmdet.core.mask.mask_target import *
>>> H, W = 17, 18
>>> cfg = mmcv.Config({'mask_size': (13, 14)})
>>> rng = np.random.RandomState(0)
>>> # Positive proposals (tl_x, tl_y, br_x, br_y) for each image
>>> pos_proposals_list = [
>>>     torch.Tensor([
>>>         [ 7.2425,  5.5929, 13.9414, 14.9541],
>>>         [ 7.3241,  3.6170, 16.3850, 15.3102],
>>>     ]),
>>>     torch.Tensor([
>>>         [ 4.8448, 6.4010, 7.0314, 9.7681],
>>>         [ 5.9790, 2.6989, 7.4416, 4.8580],
>>>         [ 0.0000, 0.0000, 0.1398, 9.8232],
>>>     ]),
>>> ]
>>> # Corresponding class index for each proposal for each image
>>> pos_assigned_gt_inds_list = [
>>>     torch.LongTensor([7, 0]),
>>>     torch.LongTensor([5, 4, 1]),
>>> ]
>>> # Ground truth mask for each true object for each image
>>> gt_masks_list = [
>>>     BitmapMasks(rng.rand(8, H, W), height=H, width=W),
>>>     BitmapMasks(rng.rand(6, H, W), height=H, width=W),
>>> ]
>>> mask_targets = mask_target(
>>>     pos_proposals_list, pos_assigned_gt_inds_list,
>>>     gt_masks_list, cfg)
>>> assert mask_targets.shape == (5,) + cfg['mask_size']
class mmdet.core.mask.BaseInstanceMasks[source]

Base class for instance masks.

areas

areas of each instance.

Type:ndarray
crop(bbox)[source]

Crop each mask by the given bbox.

Parameters:bbox (ndarray) – Bbox in format [x1, y1, x2, y2], shape (4, ).
Returns:The cropped masks.
Return type:BaseInstanceMasks
crop_and_resize(bboxes, out_shape, inds, device, interpolation='bilinear', binarize=True)[source]

Crop and resize masks by the given bboxes.

This function is mainly used in mask targets computation. It firstly align mask to bboxes by assigned_inds, then crop mask by the assigned bbox and resize to the size of (mask_h, mask_w)

Parameters:
  • bboxes (Tensor) – Bboxes in format [x1, y1, x2, y2], shape (N, 4)
  • out_shape (tuple[int]) – Target (h, w) of resized mask
  • inds (ndarray) – Indexes to assign masks to each bbox, shape (N,) and values should be between [0, num_masks - 1].
  • device (str) – Device of bboxes
  • interpolation (str) – See mmcv.imresize
  • binarize (bool) – if True fractional values are rounded to 0 or 1 after the resize operation. if False and unsupported an error will be raised. Defaults to True.
Returns:

the cropped and resized masks.

Return type:

BaseInstanceMasks

expand(expanded_h, expanded_w, top, left)[source]

see Expand.

flip(flip_direction='horizontal')[source]

Flip masks alone the given direction.

Parameters:flip_direction (str) – Either ‘horizontal’ or ‘vertical’.
Returns:The flipped masks.
Return type:BaseInstanceMasks
pad(out_shape, pad_val)[source]

Pad masks to the given size of (h, w).

Parameters:
  • out_shape (tuple[int]) – Target (h, w) of padded mask.
  • pad_val (int) – The padded value.
Returns:

The padded masks.

Return type:

BaseInstanceMasks

rescale(scale, interpolation='nearest')[source]

Rescale masks as large as possible while keeping the aspect ratio. For details can refer to mmcv.imrescale.

Parameters:
  • scale (tuple[int]) – The maximum size (h, w) of rescaled mask.
  • interpolation (str) – Same as mmcv.imrescale().
Returns:

The rescaled masks.

Return type:

BaseInstanceMasks

resize(out_shape, interpolation='nearest')[source]

Resize masks to the given out_shape.

Parameters:
  • out_shape – Target (h, w) of resized mask.
  • interpolation (str) – See mmcv.imresize().
Returns:

The resized masks.

Return type:

BaseInstanceMasks

rotate(out_shape, angle, center=None, scale=1.0, fill_val=0)[source]

Rotate the masks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • angle (int | float) – Rotation angle in degrees. Positive values mean counter-clockwise rotation.
  • center (tuple[float], optional) – Center point (w, h) of the rotation in source image. If not specified, the center of the image will be used.
  • scale (int | float) – Isotropic scale factor.
  • fill_val (int | float) – Border value. Default 0 for masks.
Returns:

Rotated masks.

shear(out_shape, magnitude, direction='horizontal', border_value=0, interpolation='bilinear')[source]

Shear the masks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • magnitude (int | float) – The magnitude used for shear.
  • direction (str) – The shear direction, either “horizontal” or “vertical”.
  • border_value (int | tuple[int]) – Value used in case of a constant border. Default 0.
  • interpolation (str) – Same as in mmcv.imshear().
Returns:

Sheared masks.

Return type:

ndarray

to_ndarray()[source]

Convert masks to the format of ndarray.

Returns:Converted masks in the format of ndarray.
Return type:ndarray
to_tensor(dtype, device)[source]

Convert masks to the format of Tensor.

Parameters:
  • dtype (str) – Dtype of converted mask.
  • device (torch.device) – Device of converted masks.
Returns:

Converted masks in the format of Tensor.

Return type:

Tensor

translate(out_shape, offset, direction='horizontal', fill_val=0, interpolation='bilinear')[source]

Translate the masks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • offset (int | float) – The offset for translate.
  • direction (str) – The translate direction, either “horizontal” or “vertical”.
  • fill_val (int | float) – Border value. Default 0.
  • interpolation (str) – Same as mmcv.imtranslate().
Returns:

Translated masks.

class mmdet.core.mask.BitmapMasks(masks, height, width)[source]

This class represents masks in the form of bitmaps.

Parameters:
  • masks (ndarray) – ndarray of masks in shape (N, H, W), where N is the number of objects.
  • height (int) – height of masks
  • width (int) – width of masks

Example

>>> from mmdet.core.mask.structures import *  # NOQA
>>> num_masks, H, W = 3, 32, 32
>>> rng = np.random.RandomState(0)
>>> masks = (rng.rand(num_masks, H, W) > 0.1).astype(np.int)
>>> self = BitmapMasks(masks, height=H, width=W)
>>> # demo crop_and_resize
>>> num_boxes = 5
>>> bboxes = np.array([[0, 0, 30, 10.0]] * num_boxes)
>>> out_shape = (14, 14)
>>> inds = torch.randint(0, len(self), size=(num_boxes,))
>>> device = 'cpu'
>>> interpolation = 'bilinear'
>>> new = self.crop_and_resize(
...     bboxes, out_shape, inds, device, interpolation)
>>> assert len(new) == num_boxes
>>> assert new.height, new.width == out_shape
areas

See BaseInstanceMasks.areas.

crop(bbox)[source]

See BaseInstanceMasks.crop().

crop_and_resize(bboxes, out_shape, inds, device='cpu', interpolation='bilinear', binarize=True)[source]

See BaseInstanceMasks.crop_and_resize().

expand(expanded_h, expanded_w, top, left)[source]

See BaseInstanceMasks.expand().

flip(flip_direction='horizontal')[source]

See BaseInstanceMasks.flip().

pad(out_shape, pad_val=0)[source]

See BaseInstanceMasks.pad().

classmethod random(num_masks=3, height=32, width=32, dtype=<class 'numpy.uint8'>, rng=None)[source]

Generate random bitmap masks for demo / testing purposes.

Example

>>> from mmdet.core.mask.structures import BitmapMasks
>>> self = BitmapMasks.random()
>>> print('self = {}'.format(self))
self = BitmapMasks(num_masks=3, height=32, width=32)
rescale(scale, interpolation='nearest')[source]

See BaseInstanceMasks.rescale().

resize(out_shape, interpolation='nearest')[source]

See BaseInstanceMasks.resize().

rotate(out_shape, angle, center=None, scale=1.0, fill_val=0)[source]

Rotate the BitmapMasks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • angle (int | float) – Rotation angle in degrees. Positive values mean counter-clockwise rotation.
  • center (tuple[float], optional) – Center point (w, h) of the rotation in source image. If not specified, the center of the image will be used.
  • scale (int | float) – Isotropic scale factor.
  • fill_val (int | float) – Border value. Default 0 for masks.
Returns:

Rotated BitmapMasks.

Return type:

BitmapMasks

shear(out_shape, magnitude, direction='horizontal', border_value=0, interpolation='bilinear')[source]

Shear the BitmapMasks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • magnitude (int | float) – The magnitude used for shear.
  • direction (str) – The shear direction, either “horizontal” or “vertical”.
  • border_value (int | tuple[int]) – Value used in case of a constant border.
  • interpolation (str) – Same as in mmcv.imshear().
Returns:

The sheared masks.

Return type:

BitmapMasks

to_ndarray()[source]

See BaseInstanceMasks.to_ndarray().

to_tensor(dtype, device)[source]

See BaseInstanceMasks.to_tensor().

translate(out_shape, offset, direction='horizontal', fill_val=0, interpolation='bilinear')[source]

Translate the BitmapMasks.

Parameters:
  • out_shape (tuple[int]) – Shape for output mask, format (h, w).
  • offset (int | float) – The offset for translate.
  • direction (str) – The translate direction, either “horizontal” or “vertical”.
  • fill_val (int | float) – Border value. Default 0 for masks.
  • interpolation (str) – Same as mmcv.imtranslate().
Returns:

Translated BitmapMasks.

Return type:

BitmapMasks

Example

>>> from mmdet.core.mask.structures import BitmapMasks
>>> self = BitmapMasks.random(dtype=np.uint8)
>>> out_shape = (32, 32)
>>> offset = 4
>>> direction = 'horizontal'
>>> fill_val = 0
>>> interpolation = 'bilinear'
>>> # Note, There seem to be issues when:
>>> # * out_shape is different than self's shape
>>> # * the mask dtype is not supported by cv2.AffineWarp
>>> new = self.translate(out_shape, offset, direction, fill_val,
>>>                      interpolation)
>>> assert len(new) == len(self)
>>> assert new.height, new.width == out_shape
class mmdet.core.mask.PolygonMasks(masks, height, width)[source]

This class represents masks in the form of polygons.

Polygons is a list of three levels. The first level of the list corresponds to objects, the second level to the polys that compose the object, the third level to the poly coordinates

Parameters:
  • masks (list[list[ndarray]]) – The first level of the list corresponds to objects, the second level to the polys that compose the object, the third level to the poly coordinates
  • height (int) – height of masks
  • width (int) – width of masks

Example

>>> from mmdet.core.mask.structures import *  # NOQA
>>> masks = [
>>>     [ np.array([0, 0, 10, 0, 10, 10., 0, 10, 0, 0]) ]
>>> ]
>>> height, width = 16, 16
>>> self = PolygonMasks(masks, height, width)
>>> # demo translate
>>> new = self.translate((16, 16), 4., direction='horizontal')
>>> assert np.all(new.masks[0][0][1::2] == masks[0][0][1::2])
>>> assert np.all(new.masks[0][0][0::2] == masks[0][0][0::2] + 4)
>>> # demo crop_and_resize
>>> num_boxes = 3
>>> bboxes = np.array([[0, 0, 30, 10.0]] * num_boxes)
>>> out_shape = (16, 16)
>>> inds = torch.randint(0, len(self), size=(num_boxes,))
>>> device = 'cpu'
>>> interpolation = 'bilinear'
>>> new = self.crop_and_resize(
...     bboxes, out_shape, inds, device, interpolation)
>>> assert len(new) == num_boxes
>>> assert new.height, new.width == out_shape
areas

Compute areas of masks.

This func is modified from detectron2. The function only works with Polygons using the shoelace formula.

Returns:areas of each instance
Return type:ndarray
crop(bbox)[source]

see BaseInstanceMasks.crop()

crop_and_resize(bboxes, out_shape, inds, device='cpu', interpolation='bilinear', binarize=True)[source]

see BaseInstanceMasks.crop_and_resize()

expand(*args, **kwargs)[source]

TODO: Add expand for polygon

flip(flip_direction='horizontal')[source]

see BaseInstanceMasks.flip()

pad(out_shape, pad_val=0)[source]

padding has no effect on polygons`

classmethod random(num_masks=3, height=32, width=32, n_verts=5, dtype=<class 'numpy.float32'>, rng=None)[source]

Generate random polygon masks for demo / testing purposes.

Adapted from [1]

References

[1](1, 2) https://gitlab.kitware.com/computer-vision/kwimage/-/blob/928cae35ca8/kwimage/structs/polygon.py#L379 # noqa: E501

Example

>>> from mmdet.core.mask.structures import PolygonMasks
>>> self = PolygonMasks.random()
>>> print('self = {}'.format(self))
rescale(scale, interpolation=None)[source]

see BaseInstanceMasks.rescale()

resize(out_shape, interpolation=None)[source]

see BaseInstanceMasks.resize()

rotate(out_shape, angle, center=None, scale=1.0, fill_val=0)[source]

See BaseInstanceMasks.rotate().

shear(out_shape, magnitude, direction='horizontal', border_value=0, interpolation='bilinear')[source]

See BaseInstanceMasks.shear().

to_bitmap()[source]

convert polygon masks to bitmap masks.

to_ndarray()[source]

Convert masks to the format of ndarray.

to_tensor(dtype, device)[source]

See BaseInstanceMasks.to_tensor().

translate(out_shape, offset, direction='horizontal', fill_val=None, interpolation=None)[source]

Translate the PolygonMasks.

Example

>>> self = PolygonMasks.random(dtype=np.int)
>>> out_shape = (self.height, self.width)
>>> new = self.translate(out_shape, 4., direction='horizontal')
>>> assert np.all(new.masks[0][0][1::2] == self.masks[0][0][1::2])
>>> assert np.all(new.masks[0][0][0::2] == self.masks[0][0][0::2] + 4)  # noqa: E501
mmdet.core.mask.encode_mask_results(mask_results)[source]

Encode bitmap mask to RLE code.

Parameters:mask_results (list | tuple[list]) – bitmap mask results. In mask scoring rcnn, mask_results is a tuple of (segm_results, segm_cls_score).
Returns:RLE encoded mask.
Return type:list | tuple

evaluation

mmdet.core.evaluation.get_classes(dataset)[source]

Get class names of a dataset.

class mmdet.core.evaluation.DistEvalHook(dataloader, start=None, interval=1, by_epoch=True, save_best=None, rule=None, test_fn=None, greater_keys=None, less_keys=None, broadcast_bn_buffer=True, tmpdir=None, gpu_collect=False, **eval_kwargs)[source]
class mmdet.core.evaluation.EvalHook(dataloader, start=None, interval=1, by_epoch=True, save_best=None, rule=None, test_fn=None, greater_keys=None, less_keys=None, **eval_kwargs)[source]
mmdet.core.evaluation.average_precision(recalls, precisions, mode='area')[source]

Calculate average precision (for single or multiple scales).

Parameters:
  • recalls (ndarray) – shape (num_scales, num_dets) or (num_dets, )
  • precisions (ndarray) – shape (num_scales, num_dets) or (num_dets, )
  • mode (str) – ‘area’ or ‘11points’, ‘area’ means calculating the area under precision-recall curve, ‘11points’ means calculating the average precision of recalls at [0, 0.1, …, 1]
Returns:

calculated average precision

Return type:

float or ndarray

mmdet.core.evaluation.eval_map(det_results, annotations, scale_ranges=None, iou_thr=0.5, dataset=None, logger=None, tpfp_fn=None, nproc=4)[source]

Evaluate mAP of a dataset.

Parameters:
  • det_results (list[list]) – [[cls1_det, cls2_det, …], …]. The outer list indicates images, and the inner list indicates per-class detected bboxes.
  • annotations (list[dict]) –

    Ground truth annotations where each item of the list indicates an image. Keys of annotations are:

    • bboxes: numpy array of shape (n, 4)
    • labels: numpy array of shape (n, )
    • bboxes_ignore (optional): numpy array of shape (k, 4)
    • labels_ignore (optional): numpy array of shape (k, )
  • scale_ranges (list[tuple] | None) – Range of scales to be evaluated, in the format [(min1, max1), (min2, max2), …]. A range of (32, 64) means the area range between (32**2, 64**2). Default: None.
  • iou_thr (float) – IoU threshold to be considered as matched. Default: 0.5.
  • dataset (list[str] | str | None) – Dataset name or dataset classes, there are minor differences in metrics for different datsets, e.g. “voc07”, “imagenet_det”, etc. Default: None.
  • logger (logging.Logger | str | None) – The way to print the mAP summary. See mmcv.utils.print_log() for details. Default: None.
  • tpfp_fn (callable | None) – The function used to determine true/ false positives. If None, tpfp_default() is used as default unless dataset is ‘det’ or ‘vid’ (tpfp_imagenet() in this case). If it is given as a function, then this function is used to evaluate tp & fp. Default None.
  • nproc (int) – Processes used for computing TP and FP. Default: 4.
Returns:

(mAP, [dict, dict, …])

Return type:

tuple

mmdet.core.evaluation.print_map_summary(mean_ap, results, dataset=None, scale_ranges=None, logger=None)[source]

Print mAP and results of each class.

A table will be printed to show the gts/dets/recall/AP of each class and the mAP.

Parameters:
  • mean_ap (float) – Calculated from eval_map().
  • results (list[dict]) – Calculated from eval_map().
  • dataset (list[str] | str | None) – Dataset name or dataset classes.
  • scale_ranges (list[tuple] | None) – Range of scales to be evaluated.
  • logger (logging.Logger | str | None) – The way to print the mAP summary. See mmcv.utils.print_log() for details. Default: None.
mmdet.core.evaluation.eval_recalls(gts, proposals, proposal_nums=None, iou_thrs=0.5, logger=None)[source]

Calculate recalls.

Parameters:
  • gts (list[ndarray]) – a list of arrays of shape (n, 4)
  • proposals (list[ndarray]) – a list of arrays of shape (k, 4) or (k, 5)
  • proposal_nums (int | Sequence[int]) – Top N proposals to be evaluated.
  • iou_thrs (float | Sequence[float]) – IoU thresholds. Default: 0.5.
  • logger (logging.Logger | str | None) – The way to print the recall summary. See mmcv.utils.print_log() for details. Default: None.
Returns:

recalls of different ious and proposal nums

Return type:

ndarray

mmdet.core.evaluation.print_recall_summary(recalls, proposal_nums, iou_thrs, row_idxs=None, col_idxs=None, logger=None)[source]

Print recalls in a table.

Parameters:
  • recalls (ndarray) – calculated from bbox_recalls
  • proposal_nums (ndarray or list) – top N proposals
  • iou_thrs (ndarray or list) – iou thresholds
  • row_idxs (ndarray) – which rows(proposal nums) to print
  • col_idxs (ndarray) – which cols(iou thresholds) to print
  • logger (logging.Logger | str | None) – The way to print the recall summary. See mmcv.utils.print_log() for details. Default: None.
mmdet.core.evaluation.plot_num_recall(recalls, proposal_nums)[source]

Plot Proposal_num-Recalls curve.

Parameters:
  • recalls (ndarray or list) – shape (k,)
  • proposal_nums (ndarray or list) – same shape as recalls
mmdet.core.evaluation.plot_iou_recall(recalls, iou_thrs)[source]

Plot IoU-Recalls curve.

Parameters:
  • recalls (ndarray or list) – shape (k,)
  • iou_thrs (ndarray or list) – same shape as recalls

post_processing

mmdet.core.post_processing.multiclass_nms(multi_bboxes, multi_scores, score_thr, nms_cfg, max_num=-1, score_factors=None, return_inds=False)[source]

NMS for multi-class bboxes.

Parameters:
  • multi_bboxes (Tensor) – shape (n, #class*4) or (n, 4)
  • multi_scores (Tensor) – shape (n, #class), where the last column contains scores of the background class, but this will be ignored.
  • score_thr (float) – bbox threshold, bboxes with scores lower than it will not be considered.
  • nms_thr (float) – NMS IoU threshold
  • max_num (int, optional) – if there are more than max_num bboxes after NMS, only top max_num will be kept. Default to -1.
  • score_factors (Tensor, optional) – The factors multiplied to scores before applying NMS. Default to None.
  • return_inds (bool, optional) – Whether return the indices of kept bboxes. Default to False.
Returns:

(dets, labels, indices (optional)), tensors of shape (k, 5),

(k), and (k). Dets are boxes with scores. Labels are 0-based.

Return type:

tuple

mmdet.core.post_processing.merge_aug_proposals(aug_proposals, img_metas, cfg)[source]

Merge augmented proposals (multiscale, flip, etc.)

Parameters:
  • aug_proposals (list[Tensor]) – proposals from different testing schemes, shape (n, 5). Note that they are not rescaled to the original image size.
  • img_metas (list[dict]) – list of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet/datasets/pipelines/formatting.py:Collect.
  • cfg (dict) – rpn test config.
Returns:

shape (n, 4), proposals corresponding to original image scale.

Return type:

Tensor

mmdet.core.post_processing.merge_aug_bboxes(aug_bboxes, aug_scores, img_metas, rcnn_test_cfg)[source]

Merge augmented detection bboxes and scores.

Parameters:
  • aug_bboxes (list[Tensor]) – shape (n, 4*#class)
  • aug_scores (list[Tensor] or None) – shape (n, #class)
  • img_shapes (list[Tensor]) – shape (3, ).
  • rcnn_test_cfg (dict) – rcnn test config.
Returns:

(bboxes, scores)

Return type:

tuple

mmdet.core.post_processing.merge_aug_scores(aug_scores)[source]

Merge augmented bbox scores.

mmdet.core.post_processing.merge_aug_masks(aug_masks, img_metas, rcnn_test_cfg, weights=None)[source]

Merge augmented mask prediction.

Parameters:
  • aug_masks (list[ndarray]) – shape (n, #class, h, w)
  • img_shapes (list[ndarray]) – shape (3, ).
  • rcnn_test_cfg (dict) – rcnn test config.
Returns:

(bboxes, scores)

Return type:

tuple

mmdet.core.post_processing.fast_nms(multi_bboxes, multi_scores, multi_coeffs, score_thr, iou_thr, top_k, max_num=-1)[source]

Fast NMS in YOLACT.

Fast NMS allows already-removed detections to suppress other detections so that every instance can be decided to be kept or discarded in parallel, which is not possible in traditional NMS. This relaxation allows us to implement Fast NMS entirely in standard GPU-accelerated matrix operations.

Parameters:
  • multi_bboxes (Tensor) – shape (n, #class*4) or (n, 4)
  • multi_scores (Tensor) – shape (n, #class+1), where the last column contains scores of the background class, but this will be ignored.
  • multi_coeffs (Tensor) – shape (n, #class*coeffs_dim).
  • score_thr (float) – bbox threshold, bboxes with scores lower than it will not be considered.
  • iou_thr (float) – IoU threshold to be considered as conflicted.
  • top_k (int) – if there are more than top_k bboxes before NMS, only top top_k will be kept.
  • max_num (int) – if there are more than max_num bboxes after NMS, only top max_num will be kept. If -1, keep all the bboxes. Default: -1.
Returns:

(dets, labels, coefficients), tensors of shape (k, 5), (k, 1),

and (k, coeffs_dim). Dets are boxes with scores. Labels are 0-based.

Return type:

tuple

utils

mmdet.core.utils.allreduce_grads(params, coalesce=True, bucket_size_mb=-1)[source]

Allreduce gradients.

Parameters:
  • params (list[torch.Parameters]) – List of parameters of a model
  • coalesce (bool, optional) – Whether allreduce parameters as a whole. Defaults to True.
  • bucket_size_mb (int, optional) – Size of bucket, the unit is MB. Defaults to -1.
class mmdet.core.utils.DistOptimizerHook(*args, **kwargs)[source]

Deprecated optimizer hook for distributed training.

mmdet.core.utils.reduce_mean(tensor)[source]

“Obtain the mean of tensor on different GPUs.

mmdet.core.utils.multi_apply(func, *args, **kwargs)[source]

Apply function to a list of arguments.

Note

This function applies the func to multiple inputs and map the multiple outputs of the func into different list. Each list contains the same type of outputs corresponding to different inputs.

Parameters:func (Function) – A function that will be applied to a list of arguments
Returns:A tuple containing multiple list, each list contains a kind of returned results by the function
Return type:tuple(list)
mmdet.core.utils.unmap(data, count, inds, fill=0)[source]

Unmap a subset of item (data) back to the original set of items (of size count)

mmdet.core.utils.mask2ndarray(mask)[source]

Convert Mask to ndarray..

:param mask (BitmapMasks or PolygonMasks or: :param torch.Tensor or np.ndarray): The mask to be converted.

Returns:Ndarray mask of shape (n, h, w) that has been converted
Return type:np.ndarray
mmdet.core.utils.flip_tensor(src_tensor, flip_direction)[source]

flip tensor base on flip_direction.

Parameters:
  • src_tensor (Tensor) – input feature map, shape (B, C, H, W).
  • flip_direction (str) – The flipping direction. Options are ‘horizontal’, ‘vertical’, ‘diagonal’.
Returns:

Flipped tensor.

Return type:

out_tensor (Tensor)

mmdet.datasets

datasets

class mmdet.datasets.CustomDataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]

Custom dataset for detection.

The annotation format is shown as follows. The ann field is optional for testing.

[
    {
        'filename': 'a.jpg',
        'width': 1280,
        'height': 720,
        'ann': {
            'bboxes': <np.ndarray> (n, 4) in (x1, y1, x2, y2) order.
            'labels': <np.ndarray> (n, ),
            'bboxes_ignore': <np.ndarray> (k, 4), (optional field)
            'labels_ignore': <np.ndarray> (k, 4) (optional field)
        }
    },
    ...
]
Parameters:
  • ann_file (str) – Annotation file path.
  • pipeline (list[dict]) – Processing pipeline.
  • classes (str | Sequence[str], optional) – Specify classes to load. If is None, cls.CLASSES will be used. Default: None.
  • data_root (str, optional) – Data root for ann_file, img_prefix, seg_prefix, proposal_file if specified.
  • test_mode (bool, optional) – If set True, annotation will not be loaded.
  • filter_empty_gt (bool, optional) – If set true, images without bounding boxes of the dataset’s classes will be filtered out. This option only works when test_mode=False, i.e., we never filter images during tests.
evaluate(results, metric='mAP', logger=None, proposal_nums=(100, 300, 1000), iou_thr=0.5, scale_ranges=None)[source]

Evaluate the dataset.

Parameters:
  • results (list) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated.
  • logger (logging.Logger | None | str) – Logger used for printing related information during evaluation. Default: None.
  • proposal_nums (Sequence[int]) – Proposal number used for evaluating recalls, such as recall@100, recall@1000. Default: (100, 300, 1000).
  • iou_thr (float | list[float]) – IoU threshold. Default: 0.5.
  • scale_ranges (list[tuple] | None) – Scale ranges for evaluating mAP. Default: None.
format_results(results, **kwargs)[source]

Place holder to format result to dataset specific output.

get_ann_info(idx)[source]

Get annotation by index.

Parameters:idx (int) – Index of data.
Returns:Annotation info of specified index.
Return type:dict
get_cat_ids(idx)[source]

Get category ids by index.

Parameters:idx (int) – Index of data.
Returns:All categories in the image of specified index.
Return type:list[int]
classmethod get_classes(classes=None)[source]

Get class names of current dataset.

Parameters:classes (Sequence[str] | str | None) – If classes is None, use default CLASSES defined by builtin dataset. If classes is a string, take it as a file name. The file contains the name of classes where each line contains one class name. If classes is a tuple or list, override the CLASSES defined by the dataset.
Returns:Names of categories of the dataset.
Return type:tuple[str] or list[str]
load_annotations(ann_file)[source]

Load annotation from annotation file.

load_proposals(proposal_file)[source]

Load proposal from proposal file.

pre_pipeline(results)[source]

Prepare results dict for pipeline.

prepare_test_img(idx)[source]

Get testing data after pipeline.

Parameters:idx (int) – Index of data.
Returns:Testing data after pipeline with new keys introduced by pipeline.
Return type:dict
prepare_train_img(idx)[source]

Get training data and annotations after pipeline.

Parameters:idx (int) – Index of data.
Returns:Training data and annotation after pipeline with new keys introduced by pipeline.
Return type:dict
class mmdet.datasets.XMLDataset(min_size=None, **kwargs)[source]

XML dataset for detection.

Parameters:min_size (int | float, optional) – The minimum size of bounding boxes in the images. If the size of a bounding box is less than min_size, it would be add to ignored field.
get_ann_info(idx)[source]

Get annotation from XML file by index.

Parameters:idx (int) – Index of data.
Returns:Annotation info of specified index.
Return type:dict
get_cat_ids(idx)[source]

Get category ids in XML file by index.

Parameters:idx (int) – Index of data.
Returns:All categories in the image of specified index.
Return type:list[int]
load_annotations(ann_file)[source]

Load annotation from XML style ann_file.

Parameters:ann_file (str) – Path of XML file.
Returns:Annotation info from XML file.
Return type:list[dict]
class mmdet.datasets.CocoDataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]
evaluate(results, metric='bbox', logger=None, jsonfile_prefix=None, classwise=False, proposal_nums=(100, 300, 1000), iou_thrs=None, metric_items=None)[source]

Evaluation in COCO protocol.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated. Options are ‘bbox’, ‘segm’, ‘proposal’, ‘proposal_fast’.
  • logger (logging.Logger | str | None) – Logger used for printing related information during evaluation. Default: None.
  • jsonfile_prefix (str | None) – The prefix of json files. It includes the file path and the prefix of filename, e.g., “a/b/prefix”. If not specified, a temp file will be created. Default: None.
  • classwise (bool) – Whether to evaluating the AP for each class.
  • proposal_nums (Sequence[int]) – Proposal number used for evaluating recalls, such as recall@100, recall@1000. Default: (100, 300, 1000).
  • iou_thrs (Sequence[float], optional) – IoU threshold used for evaluating recalls/mAPs. If set to a list, the average of all IoUs will also be computed. If not specified, [0.50, 0.55, 0.60, 0.65, 0.70, 0.75, 0.80, 0.85, 0.90, 0.95] will be used. Default: None.
  • metric_items (list[str] | str, optional) – Metric items that will be returned. If not specified, ['AR@100', 'AR@300', 'AR@1000', 'AR_s@1000', 'AR_m@1000', 'AR_l@1000' ] will be used when metric=='proposal', ['mAP', 'mAP_50', 'mAP_75', 'mAP_s', 'mAP_m', 'mAP_l'] will be used when metric=='bbox' or metric=='segm'.
Returns:

COCO style evaluation metric.

Return type:

dict[str, float]

format_results(results, jsonfile_prefix=None, **kwargs)[source]

Format the results to json (standard format for COCO evaluation).

Parameters:
  • results (list[tuple | numpy.ndarray]) – Testing results of the dataset.
  • jsonfile_prefix (str | None) – The prefix of json files. It includes the file path and the prefix of filename, e.g., “a/b/prefix”. If not specified, a temp file will be created. Default: None.
Returns:

(result_files, tmp_dir), result_files is a dict containing the json filepaths, tmp_dir is the temporal directory created for saving json files when jsonfile_prefix is not specified.

Return type:

tuple

get_ann_info(idx)[source]

Get COCO annotation by index.

Parameters:idx (int) – Index of data.
Returns:Annotation info of specified index.
Return type:dict
get_cat_ids(idx)[source]

Get COCO category ids by index.

Parameters:idx (int) – Index of data.
Returns:All categories in the image of specified index.
Return type:list[int]
load_annotations(ann_file)[source]

Load annotation from COCO style annotation file.

Parameters:ann_file (str) – Path of annotation file.
Returns:Annotation info from COCO api.
Return type:list[dict]
results2json(results, outfile_prefix)[source]

Dump the detection results to a COCO style json file.

There are 3 types of results: proposals, bbox predictions, mask predictions, and they have different data types. This method will automatically recognize the type, and dump them to json files.

Parameters:
  • results (list[list | tuple | ndarray]) – Testing results of the dataset.
  • outfile_prefix (str) – The filename prefix of the json files. If the prefix is “somepath/xxx”, the json files will be named “somepath/xxx.bbox.json”, “somepath/xxx.segm.json”, “somepath/xxx.proposal.json”.
Returns:

str]: Possible keys are “bbox”, “segm”, “proposal”, and values are corresponding filenames.

Return type:

dict[str

xyxy2xywh(bbox)[source]

Convert xyxy style bounding boxes to xywh style for COCO evaluation.

Parameters:bbox (numpy.ndarray) – The bounding boxes, shape (4, ), in xyxy order.
Returns:The converted bounding boxes, in xywh order.
Return type:list[float]
class mmdet.datasets.DeepFashionDataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]
class mmdet.datasets.VOCDataset(**kwargs)[source]
evaluate(results, metric='mAP', logger=None, proposal_nums=(100, 300, 1000), iou_thr=0.5, scale_ranges=None)[source]

Evaluate in VOC protocol.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated. Options are ‘mAP’, ‘recall’.
  • logger (logging.Logger | str, optional) – Logger used for printing related information during evaluation. Default: None.
  • proposal_nums (Sequence[int]) – Proposal number used for evaluating recalls, such as recall@100, recall@1000. Default: (100, 300, 1000).
  • iou_thr (float | list[float]) – IoU threshold. Default: 0.5.
  • scale_ranges (list[tuple], optional) – Scale ranges for evaluating mAP. If not specified, all bounding boxes would be included in evaluation. Default: None.
Returns:

AP/recall metrics.

Return type:

dict[str, float]

class mmdet.datasets.CityscapesDataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]
evaluate(results, metric='bbox', logger=None, outfile_prefix=None, classwise=False, proposal_nums=(100, 300, 1000), iou_thrs=array([0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95]))[source]

Evaluation in Cityscapes/COCO protocol.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated. Options are ‘bbox’, ‘segm’, ‘proposal’, ‘proposal_fast’.
  • logger (logging.Logger | str | None) – Logger used for printing related information during evaluation. Default: None.
  • outfile_prefix (str | None) – The prefix of output file. It includes the file path and the prefix of filename, e.g., “a/b/prefix”. If results are evaluated with COCO protocol, it would be the prefix of output json file. For example, the metric is ‘bbox’ and ‘segm’, then json files would be “a/b/prefix.bbox.json” and “a/b/prefix.segm.json”. If results are evaluated with cityscapes protocol, it would be the prefix of output txt/png files. The output files would be png images under folder “a/b/prefix/xxx/” and the file name of images would be written into a txt file “a/b/prefix/xxx_pred.txt”, where “xxx” is the video name of cityscapes. If not specified, a temp file will be created. Default: None.
  • classwise (bool) – Whether to evaluating the AP for each class.
  • proposal_nums (Sequence[int]) – Proposal number used for evaluating recalls, such as recall@100, recall@1000. Default: (100, 300, 1000).
  • iou_thrs (Sequence[float]) – IoU threshold used for evaluating recalls. If set to a list, the average recall of all IoUs will also be computed. Default: 0.5.
Returns:

COCO style evaluation metric or cityscapes mAP and AP@50.

Return type:

dict[str, float]

format_results(results, txtfile_prefix=None)[source]

Format the results to txt (standard format for Cityscapes evaluation).

Parameters:
  • results (list) – Testing results of the dataset.
  • txtfile_prefix (str | None) – The prefix of txt files. It includes the file path and the prefix of filename, e.g., “a/b/prefix”. If not specified, a temp file will be created. Default: None.
Returns:

(result_files, tmp_dir), result_files is a dict containing the json filepaths, tmp_dir is the temporal directory created for saving txt/png files when txtfile_prefix is not specified.

Return type:

tuple

results2txt(results, outfile_prefix)[source]

Dump the detection results to a txt file.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • outfile_prefix (str) – The filename prefix of the json files. If the prefix is “somepath/xxx”, the txt files will be named “somepath/xxx.txt”.
Returns:

Result txt files which contains corresponding instance segmentation images.

Return type:

list[str]

mmdet.datasets.LVISDataset

alias of mmdet.datasets.lvis.LVISV05Dataset

class mmdet.datasets.LVISV05Dataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]
evaluate(results, metric='bbox', logger=None, jsonfile_prefix=None, classwise=False, proposal_nums=(100, 300, 1000), iou_thrs=array([0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95]))[source]

Evaluation in LVIS protocol.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated. Options are ‘bbox’, ‘segm’, ‘proposal’, ‘proposal_fast’.
  • logger (logging.Logger | str | None) – Logger used for printing related information during evaluation. Default: None.
  • jsonfile_prefix (str | None) –
  • classwise (bool) – Whether to evaluating the AP for each class.
  • proposal_nums (Sequence[int]) – Proposal number used for evaluating recalls, such as recall@100, recall@1000. Default: (100, 300, 1000).
  • iou_thrs (Sequence[float]) – IoU threshold used for evaluating recalls. If set to a list, the average recall of all IoUs will also be computed. Default: 0.5.
Returns:

LVIS style metrics.

Return type:

dict[str, float]

load_annotations(ann_file)[source]

Load annotation from lvis style annotation file.

Parameters:ann_file (str) – Path of annotation file.
Returns:Annotation info from LVIS api.
Return type:list[dict]
class mmdet.datasets.LVISV1Dataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]
load_annotations(ann_file)[source]

Load annotation from lvis style annotation file.

Parameters:ann_file (str) – Path of annotation file.
Returns:Annotation info from LVIS api.
Return type:list[dict]
class mmdet.datasets.GroupSampler(dataset, samples_per_gpu=1)[source]
class mmdet.datasets.DistributedGroupSampler(dataset, samples_per_gpu=1, num_replicas=None, rank=None, seed=0)[source]

Sampler that restricts data loading to a subset of the dataset.

It is especially useful in conjunction with torch.nn.parallel.DistributedDataParallel. In such case, each process can pass a DistributedSampler instance as a DataLoader sampler, and load a subset of the original dataset that is exclusive to it.

Note

Dataset is assumed to be of constant size.

Parameters:
  • dataset – Dataset used for sampling.
  • num_replicas (optional) – Number of processes participating in distributed training.
  • rank (optional) – Rank of the current process within num_replicas.
  • seed (int, optional) – random seed used to shuffle the sampler if shuffle=True. This number should be identical across all processes in the distributed group. Default: 0.
class mmdet.datasets.DistributedSampler(dataset, num_replicas=None, rank=None, shuffle=True, seed=0)[source]
mmdet.datasets.build_dataloader(dataset, samples_per_gpu, workers_per_gpu, num_gpus=1, dist=True, shuffle=True, seed=None, **kwargs)[source]

Build PyTorch DataLoader.

In distributed training, each GPU/process has a dataloader. In non-distributed training, there is only one dataloader for all GPUs.

Parameters:
  • dataset (Dataset) – A PyTorch dataset.
  • samples_per_gpu (int) – Number of training samples on each GPU, i.e., batch size of each GPU.
  • workers_per_gpu (int) – How many subprocesses to use for data loading for each GPU.
  • num_gpus (int) – Number of GPUs. Only used in non-distributed training.
  • dist (bool) – Distributed training/test or not. Default: True.
  • shuffle (bool) – Whether to shuffle the data at every epoch. Default: True.
  • kwargs – any keyword argument to be used to initialize DataLoader
Returns:

A PyTorch dataloader.

Return type:

DataLoader

class mmdet.datasets.ConcatDataset(datasets, separate_eval=True)[source]

A wrapper of concatenated dataset.

Same as torch.utils.data.dataset.ConcatDataset, but concat the group flag for image aspect ratio.

Parameters:
  • datasets (list[Dataset]) – A list of datasets.
  • separate_eval (bool) – Whether to evaluate the results separately if it is used as validation dataset. Defaults to True.
evaluate(results, logger=None, **kwargs)[source]

Evaluate the results.

Parameters:
  • results (list[list | tuple]) – Testing results of the dataset.
  • logger (logging.Logger | str | None) – Logger used for printing related information during evaluation. Default: None.
Returns:

float]: AP results of the total dataset or each separate dataset if self.separate_eval=True.

Return type:

dict[str

get_cat_ids(idx)[source]

Get category ids of concatenated dataset by index.

Parameters:idx (int) – Index of data.
Returns:All categories in the image of specified index.
Return type:list[int]
class mmdet.datasets.RepeatDataset(dataset, times)[source]

A wrapper of repeated dataset.

The length of repeated dataset will be times larger than the original dataset. This is useful when the data loading time is long but the dataset is small. Using RepeatDataset can reduce the data loading time between epochs.

Parameters:
  • dataset (Dataset) – The dataset to be repeated.
  • times (int) – Repeat times.
get_cat_ids(idx)[source]

Get category ids of repeat dataset by index.

Parameters:idx (int) – Index of data.
Returns:All categories in the image of specified index.
Return type:list[int]
class mmdet.datasets.ClassBalancedDataset(dataset, oversample_thr, filter_empty_gt=True)[source]

A wrapper of repeated dataset with repeat factor.

Suitable for training on class imbalanced datasets like LVIS. Following the sampling strategy in the paper, in each epoch, an image may appear multiple times based on its “repeat factor”. The repeat factor for an image is a function of the frequency the rarest category labeled in that image. The “frequency of category c” in [0, 1] is defined by the fraction of images in the training set (without repeats) in which category c appears. The dataset needs to instantiate self.get_cat_ids() to support ClassBalancedDataset.

The repeat factor is computed as followed.

  1. For each category c, compute the fraction # of images that contain it: \(f(c)\)
  2. For each category c, compute the category-level repeat factor: \(r(c) = max(1, sqrt(t/f(c)))\)
  3. For each image I, compute the image-level repeat factor: \(r(I) = max_{c in I} r(c)\)
Parameters:
  • dataset (CustomDataset) – The dataset to be repeated.
  • oversample_thr (float) – frequency threshold below which data is repeated. For categories with f_c >= oversample_thr, there is no oversampling. For categories with f_c < oversample_thr, the degree of oversampling following the square-root inverse frequency heuristic above.
  • filter_empty_gt (bool, optional) – If set true, images without bounding boxes will not be oversampled. Otherwise, they will be categorized as the pure background class and involved into the oversampling. Default: True.
class mmdet.datasets.WIDERFaceDataset(**kwargs)[source]

Reader for the WIDER Face dataset in PASCAL VOC format.

Conversion scripts can be found in https://github.com/sovrasov/wider-face-pascal-voc-annotations

load_annotations(ann_file)[source]

Load annotation from WIDERFace XML style annotation file.

Parameters:ann_file (str) – Path of XML file.
Returns:Annotation info from XML file.
Return type:list[dict]
mmdet.datasets.replace_ImageToTensor(pipelines)[source]

Replace the ImageToTensor transform in a data pipeline to DefaultFormatBundle, which is normally useful in batch inference.

Parameters:pipelines (list[dict]) – Data pipeline configs.
Returns:
The new pipeline list with all ImageToTensor replaced by
DefaultFormatBundle.
Return type:list

Examples

>>> pipelines = [
...    dict(type='LoadImageFromFile'),
...    dict(
...        type='MultiScaleFlipAug',
...        img_scale=(1333, 800),
...        flip=False,
...        transforms=[
...            dict(type='Resize', keep_ratio=True),
...            dict(type='RandomFlip'),
...            dict(type='Normalize', mean=[0, 0, 0], std=[1, 1, 1]),
...            dict(type='Pad', size_divisor=32),
...            dict(type='ImageToTensor', keys=['img']),
...            dict(type='Collect', keys=['img']),
...        ])
...    ]
>>> expected_pipelines = [
...    dict(type='LoadImageFromFile'),
...    dict(
...        type='MultiScaleFlipAug',
...        img_scale=(1333, 800),
...        flip=False,
...        transforms=[
...            dict(type='Resize', keep_ratio=True),
...            dict(type='RandomFlip'),
...            dict(type='Normalize', mean=[0, 0, 0], std=[1, 1, 1]),
...            dict(type='Pad', size_divisor=32),
...            dict(type='DefaultFormatBundle'),
...            dict(type='Collect', keys=['img']),
...        ])
...    ]
>>> assert expected_pipelines == replace_ImageToTensor(pipelines)
mmdet.datasets.get_loading_pipeline(pipeline)[source]

Only keep loading image and annotations related configuration.

Parameters:pipeline (list[dict]) – Data pipeline configs.
Returns:
The new pipeline list with only keep
loading image and annotations related configuration.
Return type:list[dict]

Examples

>>> pipelines = [
...    dict(type='LoadImageFromFile'),
...    dict(type='LoadAnnotations', with_bbox=True),
...    dict(type='Resize', img_scale=(1333, 800), keep_ratio=True),
...    dict(type='RandomFlip', flip_ratio=0.5),
...    dict(type='Normalize', **img_norm_cfg),
...    dict(type='Pad', size_divisor=32),
...    dict(type='DefaultFormatBundle'),
...    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels'])
...    ]
>>> expected_pipelines = [
...    dict(type='LoadImageFromFile'),
...    dict(type='LoadAnnotations', with_bbox=True)
...    ]
>>> assert expected_pipelines ==        ...        get_loading_pipeline(pipelines)
class mmdet.datasets.CocoPanopticDataset(ann_file, pipeline, classes=None, data_root=None, img_prefix='', seg_prefix=None, proposal_file=None, test_mode=False, filter_empty_gt=True)[source]

Coco dataset for Panoptic segmentation.

The annotation format is shown as follows. The ann field is optional for testing.

[
    {
        'filename': f'{image_id:012}.png',
        'image_id':9
        'segments_info': {
            [
                {
                    'id': 8345037, (segment_id in panoptic png,
                                    convert from rgb)
                    'category_id': 51,
                    'iscrowd': 0,
                    'bbox': (x1, y1, w, h),
                    'area': 24315,
                    'segmentation': list,(coded mask)
                },
                ...
            }
        }
    },
    ...
]
evaluate(results, metric='pq', logger=None, jsonfile_prefix=None, **kwargs)[source]

Evaluation in COCO Panoptic protocol.

Parameters:
  • results (list[dict]) – Testing results of the dataset.
  • metric (str | list[str]) – Metrics to be evaluated. Only support ‘pq’ at present.
  • logger (logging.Logger | str | None) – Logger used for printing related information during evaluation. Default: None.
  • jsonfile_prefix (str | None) – The prefix of json files. It includes the file path and the prefix of filename, e.g., “a/b/prefix”. If not specified, a temp file will be created. Default: None.
Returns:

COCO Panoptic style evaluation metric.

Return type:

dict[str, float]

evaluate_pan_json(result_files, outfile_prefix, logger=None)[source]

Evaluate PQ according to the panoptic results json file.

get_ann_info(idx)[source]

Get COCO annotation by index.

Parameters:idx (int) – Index of data.
Returns:Annotation info of specified index.
Return type:dict
load_annotations(ann_file)[source]

Load annotation from COCO Panoptic style annotation file.

Parameters:ann_file (str) – Path of annotation file.
Returns:Annotation info from COCO api.
Return type:list[dict]
results2json(results, outfile_prefix)[source]

Dump the panoptic results to a COCO panoptic style json file.

Parameters:
  • results (dict) – Testing results of the dataset.
  • outfile_prefix (str) – The filename prefix of the json files. If the prefix is “somepath/xxx”, the json files will be named “somepath/xxx.panoptic.json”
Returns:

str]: The key is ‘panoptic’ and the value is

corresponding filename.

Return type:

dict[str

pipelines

class mmdet.datasets.pipelines.Compose(transforms)[source]

Compose multiple transforms sequentially.

Parameters:transforms (Sequence[dict | callable]) – Sequence of transform object or config dict to be composed.
mmdet.datasets.pipelines.to_tensor(data)[source]

Convert objects of various python types to torch.Tensor.

Supported types are: numpy.ndarray, torch.Tensor, Sequence, int and float.

Parameters:data (torch.Tensor | numpy.ndarray | Sequence | int | float) – Data to be converted.
class mmdet.datasets.pipelines.ToTensor(keys)[source]

Convert some results to torch.Tensor by given keys.

Parameters:keys (Sequence[str]) – Keys that need to be converted to Tensor.
class mmdet.datasets.pipelines.ImageToTensor(keys)[source]

Convert image to torch.Tensor by given keys.

The dimension order of input image is (H, W, C). The pipeline will convert it to (C, H, W). If only 2 dimension (H, W) is given, the output would be (1, H, W).

Parameters:keys (Sequence[str]) – Key of images to be converted to Tensor.
class mmdet.datasets.pipelines.ToDataContainer(fields=({'key': 'img', 'stack': True}, {'key': 'gt_bboxes'}, {'key': 'gt_labels'}))[source]

Convert results to mmcv.DataContainer by given fields.

Parameters:fields (Sequence[dict]) – Each field is a dict like dict(key='xxx', **kwargs). The key in result will be converted to mmcv.DataContainer with **kwargs. Default: (dict(key='img', stack=True), dict(key='gt_bboxes'), dict(key='gt_labels')).
class mmdet.datasets.pipelines.Transpose(keys, order)[source]

Transpose some results by given keys.

Parameters:
  • keys (Sequence[str]) – Keys of results to be transposed.
  • order (Sequence[int]) – Order of transpose.
class mmdet.datasets.pipelines.Collect(keys, meta_keys=('filename', 'ori_filename', 'ori_shape', 'img_shape', 'pad_shape', 'scale_factor', 'flip', 'flip_direction', 'img_norm_cfg'))[source]

Collect data from the loader relevant to the specific task.

This is usually the last stage of the data loader pipeline. Typically keys is set to some subset of “img”, “proposals”, “gt_bboxes”, “gt_bboxes_ignore”, “gt_labels”, and/or “gt_masks”.

The “img_meta” item is always populated. The contents of the “img_meta” dictionary depends on “meta_keys”. By default this includes:

  • “img_shape”: shape of the image input to the network as a tuple (h, w, c). Note that images may be zero padded on the bottom/right if the batch tensor is larger than this shape.

  • “scale_factor”: a float indicating the preprocessing scale

  • “flip”: a boolean indicating if image flip transform was used

  • “filename”: path to the image file

  • “ori_shape”: original shape of the image as a tuple (h, w, c)

  • “pad_shape”: image shape after padding

  • “img_norm_cfg”: a dict of normalization information:

    • mean - per channel mean subtraction
    • std - per channel std divisor
    • to_rgb - bool indicating if bgr was converted to rgb
Parameters:
  • keys (Sequence[str]) – Keys of results to be collected in data.
  • meta_keys (Sequence[str], optional) – Meta keys to be converted to mmcv.DataContainer and collected in data[img_metas]. Default: ('filename', 'ori_filename', 'ori_shape', 'img_shape', 'pad_shape', 'scale_factor', 'flip', 'flip_direction', 'img_norm_cfg')
class mmdet.datasets.pipelines.DefaultFormatBundle[source]

Default formatting bundle.

It simplifies the pipeline of formatting common fields, including “img”, “proposals”, “gt_bboxes”, “gt_labels”, “gt_masks” and “gt_semantic_seg”. These fields are formatted as follows.

  • img: (1)transpose, (2)to tensor, (3)to DataContainer (stack=True)
  • proposals: (1)to tensor, (2)to DataContainer
  • gt_bboxes: (1)to tensor, (2)to DataContainer
  • gt_bboxes_ignore: (1)to tensor, (2)to DataContainer
  • gt_labels: (1)to tensor, (2)to DataContainer
  • gt_masks: (1)to tensor, (2)to DataContainer (cpu_only=True)
  • gt_semantic_seg: (1)unsqueeze dim-0 (2)to tensor, (3)to DataContainer (stack=True)
class mmdet.datasets.pipelines.LoadAnnotations(with_bbox=True, with_label=True, with_mask=False, with_seg=False, poly2mask=True, file_client_args={'backend': 'disk'})[source]

Load multiple types of annotations.

Parameters:
  • with_bbox (bool) – Whether to parse and load the bbox annotation. Default: True.
  • with_label (bool) – Whether to parse and load the label annotation. Default: True.
  • with_mask (bool) – Whether to parse and load the mask annotation. Default: False.
  • with_seg (bool) – Whether to parse and load the semantic segmentation annotation. Default: False.
  • poly2mask (bool) – Whether to convert the instance masks from polygons to bitmaps. Default: True.
  • file_client_args (dict) – Arguments to instantiate a FileClient. See mmcv.fileio.FileClient for details. Defaults to dict(backend='disk').
process_polygons(polygons)[source]

Convert polygons to list of ndarray and filter invalid polygons.

Parameters:polygons (list[list]) – Polygons of one instance.
Returns:Processed polygons.
Return type:list[numpy.ndarray]
class mmdet.datasets.pipelines.LoadImageFromFile(to_float32=False, color_type='color', file_client_args={'backend': 'disk'})[source]

Load an image from file.

Required keys are “img_prefix” and “img_info” (a dict that must contain the key “filename”). Added or updated keys are “filename”, “img”, “img_shape”, “ori_shape” (same as img_shape), “pad_shape” (same as img_shape), “scale_factor” (1.0) and “img_norm_cfg” (means=0 and stds=1).

Parameters:
  • to_float32 (bool) – Whether to convert the loaded image to a float32 numpy array. If set to False, the loaded image is an uint8 array. Defaults to False.
  • color_type (str) – The flag argument for mmcv.imfrombytes(). Defaults to ‘color’.
  • file_client_args (dict) – Arguments to instantiate a FileClient. See mmcv.fileio.FileClient for details. Defaults to dict(backend='disk').
class mmdet.datasets.pipelines.LoadImageFromWebcam(to_float32=False, color_type='color', file_client_args={'backend': 'disk'})[source]

Load an image from webcam.

Similar with LoadImageFromFile, but the image read from webcam is in results['img'].

class mmdet.datasets.pipelines.LoadMultiChannelImageFromFiles(to_float32=False, color_type='unchanged', file_client_args={'backend': 'disk'})[source]

Load multi-channel images from a list of separate channel files.

Required keys are “img_prefix” and “img_info” (a dict that must contain the key “filename”, which is expected to be a list of filenames). Added or updated keys are “filename”, “img”, “img_shape”, “ori_shape” (same as img_shape), “pad_shape” (same as img_shape), “scale_factor” (1.0) and “img_norm_cfg” (means=0 and stds=1).

Parameters:
  • to_float32 (bool) – Whether to convert the loaded image to a float32 numpy array. If set to False, the loaded image is an uint8 array. Defaults to False.
  • color_type (str) – The flag argument for mmcv.imfrombytes(). Defaults to ‘color’.
  • file_client_args (dict) – Arguments to instantiate a FileClient. See mmcv.fileio.FileClient for details. Defaults to dict(backend='disk').
class mmdet.datasets.pipelines.LoadProposals(num_max_proposals=None)[source]

Load proposal pipeline.

Required key is “proposals”. Updated keys are “proposals”, “bbox_fields”.

Parameters:num_max_proposals (int, optional) – Maximum number of proposals to load. If not specified, all proposals will be loaded.
class mmdet.datasets.pipelines.MultiScaleFlipAug(transforms, img_scale=None, scale_factor=None, flip=False, flip_direction='horizontal')[source]

Test-time augmentation with multiple scales and flipping.

An example configuration is as followed:

After MultiScaleFLipAug with above configuration, the results are wrapped into lists of the same length as followed:

Parameters:
  • transforms (list[dict]) – Transforms to apply in each augmentation.
  • img_scale (tuple | list[tuple] | None) – Images scales for resizing.
  • scale_factor (float | list[float] | None) – Scale factors for resizing.
  • flip (bool) – Whether apply flip augmentation. Default: False.
  • flip_direction (str | list[str]) – Flip augmentation directions, options are “horizontal”, “vertical” and “diagonal”. If flip_direction is a list, multiple flip augmentations will be applied. It has no effect when flip == False. Default: “horizontal”.
class mmdet.datasets.pipelines.Resize(img_scale=None, multiscale_mode='range', ratio_range=None, keep_ratio=True, bbox_clip_border=True, backend='cv2', override=False)[source]

Resize images & bbox & mask.

This transform resizes the input image to some scale. Bboxes and masks are then resized with the same scale factor. If the input dict contains the key “scale”, then the scale in the input dict is used, otherwise the specified scale in the init method is used. If the input dict contains the key “scale_factor” (if MultiScaleFlipAug does not give img_scale but scale_factor), the actual scale will be computed by image shape and scale_factor.

img_scale can either be a tuple (single-scale) or a list of tuple (multi-scale). There are 3 multiscale modes:

  • ratio_range is not None: randomly sample a ratio from the ratio range and multiply it with the image scale.
  • ratio_range is None and multiscale_mode == "range": randomly sample a scale from the multiscale range.
  • ratio_range is None and multiscale_mode == "value": randomly sample a scale from multiple scales.
Parameters:
  • img_scale (tuple or list[tuple]) – Images scales for resizing.
  • multiscale_mode (str) – Either “range” or “value”.
  • ratio_range (tuple[float]) – (min_ratio, max_ratio)
  • keep_ratio (bool) – Whether to keep the aspect ratio when resizing the image.
  • bbox_clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.
  • backend (str) – Image resize backend, choices are ‘cv2’ and ‘pillow’. These two backends generates slightly different results. Defaults to ‘cv2’.
  • override (bool, optional) – Whether to override scale and scale_factor so as to call resize twice. Default False. If True, after the first resizing, the existed scale and scale_factor will be ignored so the second resizing can be allowed. This option is a work-around for multiple times of resize in DETR. Defaults to False.
static random_sample(img_scales)[source]

Randomly sample an img_scale when multiscale_mode=='range'.

Parameters:img_scales (list[tuple]) – Images scale range for sampling. There must be two tuples in img_scales, which specify the lower and upper bound of image scales.
Returns:Returns a tuple (img_scale, None), where img_scale is sampled scale and None is just a placeholder to be consistent with random_select().
Return type:(tuple, None)
static random_sample_ratio(img_scale, ratio_range)[source]

Randomly sample an img_scale when ratio_range is specified.

A ratio will be randomly sampled from the range specified by ratio_range. Then it would be multiplied with img_scale to generate sampled scale.

Parameters:
  • img_scale (tuple) – Images scale base to multiply with ratio.
  • ratio_range (tuple[float]) – The minimum and maximum ratio to scale the img_scale.
Returns:

Returns a tuple (scale, None), where scale is sampled ratio multiplied with img_scale and None is just a placeholder to be consistent with random_select().

Return type:

(tuple, None)

static random_select(img_scales)[source]

Randomly select an img_scale from given candidates.

Parameters:img_scales (list[tuple]) – Images scales for selection.
Returns:Returns a tuple (img_scale, scale_dix), where img_scale is the selected image scale and scale_idx is the selected index in the given candidates.
Return type:(tuple, int)
class mmdet.datasets.pipelines.RandomFlip(flip_ratio=None, direction='horizontal')[source]

Flip the image & bbox & mask.

If the input dict contains the key “flip”, then the flag will be used, otherwise it will be randomly decided by a ratio specified in the init method.

When random flip is enabled, flip_ratio/direction can either be a float/string or tuple of float/string. There are 3 flip modes:

  • flip_ratio is float, direction is string: the image will be
    direction``ly flipped with probability of ``flip_ratio . E.g., flip_ratio=0.5, direction='horizontal', then image will be horizontally flipped with probability of 0.5.
  • flip_ratio is float, direction is list of string: the image wil
    be direction[i]``ly flipped with probability of ``flip_ratio/len(direction). E.g., flip_ratio=0.5, direction=['horizontal', 'vertical'], then image will be horizontally flipped with probability of 0.25, vertically with probability of 0.25.
  • flip_ratio is list of float, direction is list of string:
    given len(flip_ratio) == len(direction), the image wil be direction[i]``ly flipped with probability of ``flip_ratio[i]. E.g., flip_ratio=[0.3, 0.5], direction=['horizontal', 'vertical'], then image will be horizontally flipped with probability of 0.3, vertically with probability of 0.5.
Parameters:
  • flip_ratio (float | list[float], optional) – The flipping probability. Default: None.
  • direction (str | list[str], optional) – The flipping direction. Options are ‘horizontal’, ‘vertical’, ‘diagonal’. Default: ‘horizontal’. If input is a list, the length must equal flip_ratio. Each element in flip_ratio indicates the flip probability of corresponding direction.
bbox_flip(bboxes, img_shape, direction)[source]

Flip bboxes horizontally.

Parameters:
  • bboxes (numpy.ndarray) – Bounding boxes, shape (…, 4*k)
  • img_shape (tuple[int]) – Image shape (height, width)
  • direction (str) – Flip direction. Options are ‘horizontal’, ‘vertical’.
Returns:

Flipped bounding boxes.

Return type:

numpy.ndarray

class mmdet.datasets.pipelines.Pad(size=None, size_divisor=None, pad_val=0)[source]

Pad the image & mask.

There are two padding modes: (1) pad to a fixed size and (2) pad to the minimum size that is divisible by some number. Added keys are “pad_shape”, “pad_fixed_size”, “pad_size_divisor”,

Parameters:
  • size (tuple, optional) – Fixed padding size.
  • size_divisor (int, optional) – The divisor of padded size.
  • pad_val (float, optional) – Padding value, 0 by default.
class mmdet.datasets.pipelines.RandomCrop(crop_size, crop_type='absolute', allow_negative_crop=False, bbox_clip_border=True)[source]

Random crop the image & bboxes & masks.

The absolute crop_size is sampled based on crop_type and image_size, then the cropped results are generated.

Parameters:
  • crop_size (tuple) – The relative ratio or absolute pixels of height and width.
  • crop_type (str, optional) – one of “relative_range”, “relative”, “absolute”, “absolute_range”. “relative” randomly crops (h * crop_size[0], w * crop_size[1]) part from an input of size (h, w). “relative_range” uniformly samples relative crop size from range [crop_size[0], 1] and [crop_size[1], 1] for height and width respectively. “absolute” crops from an input with absolute size (crop_size[0], crop_size[1]). “absolute_range” uniformly samples crop_h in range [crop_size[0], min(h, crop_size[1])] and crop_w in range [crop_size[0], min(w, crop_size[1])]. Default “absolute”.
  • allow_negative_crop (bool, optional) – Whether to allow a crop that does not contain any bbox area. Default False.
  • bbox_clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.

Note

  • If the image is smaller than the absolute crop size, return the
    original image.
  • The keys for bboxes, labels and masks must be aligned. That is, gt_bboxes corresponds to gt_labels and gt_masks, and gt_bboxes_ignore corresponds to gt_labels_ignore and gt_masks_ignore.
  • If the crop does not contain any gt-bbox region and allow_negative_crop is set to False, skip this image.
class mmdet.datasets.pipelines.Normalize(mean, std, to_rgb=True)[source]

Normalize the image.

Added key is “img_norm_cfg”.

Parameters:
  • mean (sequence) – Mean values of 3 channels.
  • std (sequence) – Std values of 3 channels.
  • to_rgb (bool) – Whether to convert the image from BGR to RGB, default is true.
class mmdet.datasets.pipelines.SegRescale(scale_factor=1, backend='cv2')[source]

Rescale semantic segmentation maps.

Parameters:
  • scale_factor (float) – The scale factor of the final output.
  • backend (str) – Image rescale backend, choices are ‘cv2’ and ‘pillow’. These two backends generates slightly different results. Defaults to ‘cv2’.
class mmdet.datasets.pipelines.MinIoURandomCrop(min_ious=(0.1, 0.3, 0.5, 0.7, 0.9), min_crop_size=0.3, bbox_clip_border=True)[source]

Random crop the image & bboxes, the cropped patches have minimum IoU requirement with original image & bboxes, the IoU threshold is randomly selected from min_ious.

Parameters:
  • min_ious (tuple) – minimum IoU threshold for all intersections with
  • boxes (bounding) –
  • min_crop_size (float) – minimum crop’s size (i.e. h,w := a*h, a*w,
  • a >= min_crop_size) (where) –
  • bbox_clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.

Note

The keys for bboxes, labels and masks should be paired. That is, gt_bboxes corresponds to gt_labels and gt_masks, and gt_bboxes_ignore to gt_labels_ignore and gt_masks_ignore.

class mmdet.datasets.pipelines.Expand(mean=(0, 0, 0), to_rgb=True, ratio_range=(1, 4), seg_ignore_label=None, prob=0.5)[source]

Random expand the image & bboxes.

Randomly place the original image on a canvas of ‘ratio’ x original image size filled with mean values. The ratio is in the range of ratio_range.

Parameters:
  • mean (tuple) – mean value of dataset.
  • to_rgb (bool) – if need to convert the order of mean to align with RGB.
  • ratio_range (tuple) – range of expand ratio.
  • prob (float) – probability of applying this transformation
class mmdet.datasets.pipelines.PhotoMetricDistortion(brightness_delta=32, contrast_range=(0.5, 1.5), saturation_range=(0.5, 1.5), hue_delta=18)[source]

Apply photometric distortion to image sequentially, every transformation is applied with a probability of 0.5. The position of random contrast is in second or second to last.

  1. random brightness
  2. random contrast (mode 0)
  3. convert color from BGR to HSV
  4. random saturation
  5. random hue
  6. convert color from HSV to BGR
  7. random contrast (mode 1)
  8. randomly swap channels
Parameters:
  • brightness_delta (int) – delta of brightness.
  • contrast_range (tuple) – range of contrast.
  • saturation_range (tuple) – range of saturation.
  • hue_delta (int) – delta of hue.
class mmdet.datasets.pipelines.Albu(transforms, bbox_params=None, keymap=None, update_pad_shape=False, skip_img_without_anno=False)[source]

Albumentation augmentation.

Adds custom transformations from Albumentations library. Please, visit https://albumentations.readthedocs.io to get more information.

An example of transforms is as followed:

Parameters:
  • transforms (list[dict]) – A list of albu transformations
  • bbox_params (dict) – Bbox_params for albumentation Compose
  • keymap (dict) – Contains {‘input key’:’albumentation-style key’}
  • skip_img_without_anno (bool) – Whether to skip the image if no ann left after aug
albu_builder(cfg)[source]

Import a module from albumentations.

It inherits some of build_from_cfg() logic.

Parameters:cfg (dict) – Config dict. It should at least contain the key “type”.
Returns:The constructed object.
Return type:obj
static mapper(d, keymap)[source]

Dictionary mapper. Renames keys according to keymap provided.

Parameters:
  • d (dict) – old dict
  • keymap (dict) – {‘old_key’:’new_key’}
Returns:

new dict.

Return type:

dict

class mmdet.datasets.pipelines.InstaBoost(action_candidate=('normal', 'horizontal', 'skip'), action_prob=(1, 0, 0), scale=(0.8, 1.2), dx=15, dy=15, theta=(-1, 1), color_prob=0.5, hflag=False, aug_ratio=0.5)[source]

Data augmentation method in InstaBoost: Boosting Instance Segmentation Via Probability Map Guided Copy-Pasting.

Refer to https://github.com/GothicAi/Instaboost for implementation details.

Parameters:
  • action_candidate (tuple) – Action candidates. “normal”, “horizontal”, “vertical”, “skip” are supported. Default: (‘normal’, ‘horizontal’, ‘skip’).
  • action_prob (tuple) – Corresponding action probabilities. Should be the same length as action_candidate. Default: (1, 0, 0).
  • scale (tuple) – (min scale, max scale). Default: (0.8, 1.2).
  • dx (int) – The maximum x-axis shift will be (instance width) / dx. Default 15.
  • dy (int) – The maximum y-axis shift will be (instance height) / dy. Default 15.
  • theta (tuple) – (min rotation degree, max rotation degree). Default: (-1, 1).
  • color_prob (float) – Probability of images for color augmentation. Default 0.5.
  • heatmap_flag (bool) – Whether to use heatmap guided. Default False.
  • aug_ratio (float) – Probability of applying this transformation. Default 0.5.
class mmdet.datasets.pipelines.RandomCenterCropPad(crop_size=None, ratios=(0.9, 1.0, 1.1), border=128, mean=None, std=None, to_rgb=None, test_mode=False, test_pad_mode=('logical_or', 127), test_pad_add_pix=0, bbox_clip_border=True)[source]

Random center crop and random around padding for CornerNet.

This operation generates randomly cropped image from the original image and pads it simultaneously. Different from RandomCrop, the output shape may not equal to crop_size strictly. We choose a random value from ratios and the output shape could be larger or smaller than crop_size. The padding operation is also different from Pad, here we use around padding instead of right-bottom padding.

The relation between output image (padding image) and original image:

                output image

       +----------------------------+
       |          padded area       |
+------|----------------------------|----------+
|      |         cropped area       |          |
|      |         +---------------+  |          |
|      |         |    .   center |  |          | original image
|      |         |        range  |  |          |
|      |         +---------------+  |          |
+------|----------------------------|----------+
       |          padded area       |
       +----------------------------+

There are 5 main areas in the figure:

  • output image: output image of this operation, also called padding image in following instruction.
  • original image: input image of this operation.
  • padded area: non-intersect area of output image and original image.
  • cropped area: the overlap of output image and original image.
  • center range: a smaller area where random center chosen from. center range is computed by border and original image’s shape to avoid our random center is too close to original image’s border.

Also this operation act differently in train and test mode, the summary pipeline is listed below.

Train pipeline:

  1. Choose a random_ratio from ratios, the shape of padding image will be random_ratio * crop_size.
  2. Choose a random_center in center range.
  3. Generate padding image with center matches the random_center.
  4. Initialize the padding image with pixel value equals to mean.
  5. Copy the cropped area to padding image.
  6. Refine annotations.

Test pipeline:

  1. Compute output shape according to test_pad_mode.
  2. Generate padding image with center matches the original image center.
  3. Initialize the padding image with pixel value equals to mean.
  4. Copy the cropped area to padding image.
Parameters:
  • crop_size (tuple | None) – expected size after crop, final size will computed according to ratio. Requires (h, w) in train mode, and None in test mode.
  • ratios (tuple) – random select a ratio from tuple and crop image to (crop_size[0] * ratio) * (crop_size[1] * ratio). Only available in train mode.
  • border (int) – max distance from center select area to image border. Only available in train mode.
  • mean (sequence) – Mean values of 3 channels.
  • std (sequence) – Std values of 3 channels.
  • to_rgb (bool) – Whether to convert the image from BGR to RGB.
  • test_mode (bool) – whether involve random variables in transform. In train mode, crop_size is fixed, center coords and ratio is random selected from predefined lists. In test mode, crop_size is image’s original shape, center coords and ratio is fixed.
  • test_pad_mode (tuple) –

    padding method and padding shape value, only available in test mode. Default is using ‘logical_or’ with 127 as padding shape value.

    • ’logical_or’: final_shape = input_shape | padding_shape_value
    • ’size_divisor’: final_shape = int( ceil(input_shape / padding_shape_value) * padding_shape_value)
  • test_pad_add_pix (int) – Extra padding pixel in test mode. Default 0.
  • bbox_clip_border (bool, optional) – Whether clip the objects outside the border of the image. Defaults to True.
class mmdet.datasets.pipelines.AutoAugment(policies)[source]

Auto augmentation.

This data augmentation is proposed in Learning Data Augmentation Strategies for Object Detection.

TODO: Implement ‘Shear’, ‘Sharpness’ and ‘Rotate’ transforms

Parameters:policies (list[list[dict]]) – The policies of auto augmentation. Each policy in policies is a specific augmentation policy, and is composed by several augmentations (dict). When AutoAugment is called, a random policy in policies will be selected to augment images.

Examples

>>> replace = (104, 116, 124)
>>> policies = [
>>>     [
>>>         dict(type='Sharpness', prob=0.0, level=8),
>>>         dict(
>>>             type='Shear',
>>>             prob=0.4,
>>>             level=0,
>>>             replace=replace,
>>>             axis='x')
>>>     ],
>>>     [
>>>         dict(
>>>             type='Rotate',
>>>             prob=0.6,
>>>             level=10,
>>>             replace=replace),
>>>         dict(type='Color', prob=1.0, level=6)
>>>     ]
>>> ]
>>> augmentation = AutoAugment(policies)
>>> img = np.ones(100, 100, 3)
>>> gt_bboxes = np.ones(10, 4)
>>> results = dict(img=img, gt_bboxes=gt_bboxes)
>>> results = augmentation(results)
class mmdet.datasets.pipelines.CutOut(n_holes, cutout_shape=None, cutout_ratio=None, fill_in=(0, 0, 0))[source]

CutOut operation.

Randomly drop some regions of image used in Cutout.

Parameters:
  • n_holes (int | tuple[int, int]) – Number of regions to be dropped. If it is given as a list, number of holes will be randomly selected from the closed interval [n_holes[0], n_holes[1]].
  • cutout_shape (tuple[int, int] | list[tuple[int, int]]) – The candidate shape of dropped regions. It can be tuple[int, int] to use a fixed cutout shape, or list[tuple[int, int]] to randomly choose shape from the list.
  • cutout_ratio (tuple[float, float] | list[tuple[float, float]]) – The candidate ratio of dropped regions. It can be tuple[float, float] to use a fixed ratio or list[tuple[float, float]] to randomly choose ratio from the list. Please note that cutout_shape and cutout_ratio cannot be both given at the same time.
  • fill_in (tuple[float, float, float] | tuple[int, int, int]) – The value of pixel to fill in the dropped regions. Default: (0, 0, 0).
class mmdet.datasets.pipelines.Shear(level, img_fill_val=128, seg_ignore_label=255, prob=0.5, direction='horizontal', max_shear_magnitude=0.3, random_negative_prob=0.5, interpolation='bilinear')[source]

Apply Shear Transformation to image (and its corresponding bbox, mask, segmentation).

Parameters:
  • level (int | float) – The level should be in range [0,_MAX_LEVEL].
  • img_fill_val (int | float | tuple) – The filled values for image border. If float, the same fill value will be used for all the three channels of image. If tuple, the should be 3 elements.
  • seg_ignore_label (int) – The fill value used for segmentation map. Note this value must equals ignore_label in semantic_head of the corresponding config. Default 255.
  • prob (float) – The probability for performing Shear and should be in range [0, 1].
  • direction (str) – The direction for shear, either “horizontal” or “vertical”.
  • max_shear_magnitude (float) – The maximum magnitude for Shear transformation.
  • random_negative_prob (float) – The probability that turns the offset negative. Should be in range [0,1]
  • interpolation (str) – Same as in mmcv.imshear().
class mmdet.datasets.pipelines.Rotate(level, scale=1, center=None, img_fill_val=128, seg_ignore_label=255, prob=0.5, max_rotate_angle=30, random_negative_prob=0.5)[source]

Apply Rotate Transformation to image (and its corresponding bbox, mask, segmentation).

Parameters:
  • level (int | float) – The level should be in range (0,_MAX_LEVEL].
  • scale (int | float) – Isotropic scale factor. Same in mmcv.imrotate.
  • center (int | float | tuple[float]) – Center point (w, h) of the rotation in the source image. If None, the center of the image will be used. Same in mmcv.imrotate.
  • img_fill_val (int | float | tuple) – The fill value for image border. If float, the same value will be used for all the three channels of image. If tuple, the should be 3 elements (e.g. equals the number of channels for image).
  • seg_ignore_label (int) – The fill value used for segmentation map. Note this value must equals ignore_label in semantic_head of the corresponding config. Default 255.
  • prob (float) – The probability for perform transformation and should be in range 0 to 1.
  • max_rotate_angle (int | float) – The maximum angles for rotate transformation.
  • random_negative_prob (float) – The probability that turns the offset negative.
class mmdet.datasets.pipelines.ColorTransform(level, prob=0.5)[source]

Apply Color transformation to image. The bboxes, masks, and segmentations are not modified.

Parameters:
  • level (int | float) – Should be in range [0,_MAX_LEVEL].
  • prob (float) – The probability for performing Color transformation.
class mmdet.datasets.pipelines.EqualizeTransform(prob=0.5)[source]

Apply Equalize transformation to image. The bboxes, masks and segmentations are not modified.

Parameters:prob (float) – The probability for performing Equalize transformation.
class mmdet.datasets.pipelines.BrightnessTransform(level, prob=0.5)[source]

Apply Brightness transformation to image. The bboxes, masks and segmentations are not modified.

Parameters:
  • level (int | float) – Should be in range [0,_MAX_LEVEL].
  • prob (float) – The probability for performing Brightness transformation.
class mmdet.datasets.pipelines.ContrastTransform(level, prob=0.5)[source]

Apply Contrast transformation to image. The bboxes, masks and segmentations are not modified.

Parameters:
  • level (int | float) – Should be in range [0,_MAX_LEVEL].
  • prob (float) – The probability for performing Contrast transformation.
class mmdet.datasets.pipelines.Translate(level, prob=0.5, img_fill_val=128, seg_ignore_label=255, direction='horizontal', max_translate_offset=250.0, random_negative_prob=0.5, min_size=0)[source]

Translate the images, bboxes, masks and segmentation maps horizontally or vertically.

Parameters:
  • level (int | float) – The level for Translate and should be in range [0,_MAX_LEVEL].
  • prob (float) – The probability for performing translation and should be in range [0, 1].
  • img_fill_val (int | float | tuple) – The filled value for image border. If float, the same fill value will be used for all the three channels of image. If tuple, the should be 3 elements (e.g. equals the number of channels for image).
  • seg_ignore_label (int) – The fill value used for segmentation map. Note this value must equals ignore_label in semantic_head of the corresponding config. Default 255.
  • direction (str) – The translate direction, either “horizontal” or “vertical”.
  • max_translate_offset (int | float) – The maximum pixel’s offset for Translate.
  • random_negative_prob (float) – The probability that turns the offset negative.
  • min_size (int | float) – The minimum pixel for filtering invalid bboxes after the translation.
class mmdet.datasets.pipelines.RandomShift(shift_ratio=0.5, max_shift_px=32, filter_thr_px=1)[source]

Shift the image and box given shift pixels and probability.

Parameters:
  • shift_ratio (float) – Probability of shifts. Default 0.5.
  • max_shift_px (int) – The max pixels for shifting. Default 32.
  • filter_thr_px (int) – The width and height threshold for filtering. The bbox and the rest of the targets below the width and height threshold will be filtered. Default 1.

samplers

class mmdet.datasets.samplers.DistributedSampler(dataset, num_replicas=None, rank=None, shuffle=True, seed=0)[source]
class mmdet.datasets.samplers.DistributedGroupSampler(dataset, samples_per_gpu=1, num_replicas=None, rank=None, seed=0)[source]

Sampler that restricts data loading to a subset of the dataset.

It is especially useful in conjunction with torch.nn.parallel.DistributedDataParallel. In such case, each process can pass a DistributedSampler instance as a DataLoader sampler, and load a subset of the original dataset that is exclusive to it.

Note

Dataset is assumed to be of constant size.

Parameters:
  • dataset – Dataset used for sampling.
  • num_replicas (optional) – Number of processes participating in distributed training.
  • rank (optional) – Rank of the current process within num_replicas.
  • seed (int, optional) – random seed used to shuffle the sampler if shuffle=True. This number should be identical across all processes in the distributed group. Default: 0.
class mmdet.datasets.samplers.GroupSampler(dataset, samples_per_gpu=1)[source]

api_wrappers

class mmdet.datasets.api_wrappers.COCO(annotation_file=None)[source]

This class is almost the same as official pycocotools package.

It implements some snake case function aliases. So that the COCO class has the same interface as LVIS class.

mmdet.models

detectors

class mmdet.models.detectors.ATSS(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of ATSS.

class mmdet.models.detectors.BaseDetector(init_cfg=None)[source]

Base class for detectors.

aug_test(imgs, img_metas, **kwargs)[source]

Test function with test time augmentation.

extract_feat(imgs)[source]

Extract features from images.

extract_feats(imgs)[source]

Extract features from multiple images.

Parameters:imgs (list[torch.Tensor]) – A list of images. The images are augmented from the same image but in different ways.
Returns:Features of different images
Return type:list[torch.Tensor]
forward(img, img_metas, return_loss=True, **kwargs)[source]

Calls either forward_train() or forward_test() depending on whether return_loss is True.

Note this setting will change the expected inputs. When return_loss=True, img and img_meta are single-nested (i.e. Tensor and List[dict]), and when resturn_loss=False, img and img_meta should be double nested (i.e. List[Tensor], List[List[dict]]), with the outer list indicating test time augmentations.

forward_test(imgs, img_metas, **kwargs)[source]
Parameters:
  • imgs (List[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains all images in the batch.
  • img_metas (List[List[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch.
forward_train(imgs, img_metas, **kwargs)[source]
Parameters:
  • img (list[Tensor]) – List of tensors of shape (1, C, H, W). Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – List of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys, see mmdet.datasets.pipelines.Collect.
  • kwargs (keyword arguments) – Specific to concrete implementation.
show_result(img, result, score_thr=0.3, bbox_color=(72, 101, 241), text_color=(72, 101, 241), mask_color=None, thickness=2, font_size=13, win_name='', show=False, wait_time=0, out_file=None)[source]

Draw result over img.

Parameters:
  • img (str or Tensor) – The image to be displayed.
  • result (Tensor or tuple) – The results to draw over img bbox_result or (bbox_result, segm_result).
  • score_thr (float, optional) – Minimum score of bboxes to be shown. Default: 0.3.
  • bbox_color (str or tuple(int) or Color) – Color of bbox lines. The tuple of color should be in BGR order. Default: ‘green’
  • text_color (str or tuple(int) or Color) – Color of texts. The tuple of color should be in BGR order. Default: ‘green’
  • mask_color (None or str or tuple(int) or Color) – Color of masks. The tuple of color should be in BGR order. Default: None
  • thickness (int) – Thickness of lines. Default: 2
  • font_size (int) – Font size of texts. Default: 13
  • win_name (str) – The window name. Default: ‘’
  • wait_time (float) – Value of waitKey param. Default: 0.
  • show (bool) – Whether to show the image. Default: False.
  • out_file (str or None) – The filename to write the image. Default: None.
Returns:

Only if not show or out_file

Return type:

img (Tensor)

train_step(data, optimizer)[source]

The iteration step during training.

This method defines an iteration step during training, except for the back propagation and optimizer updating, which are done in an optimizer hook. Note that in some complicated cases or models, the whole process including back propagation and optimizer updating is also defined in this method, such as GAN.

Parameters:
  • data (dict) – The output of dataloader.
  • optimizer (torch.optim.Optimizer | dict) – The optimizer of runner is passed to train_step(). This argument is unused and reserved.
Returns:

It should contain at least 3 keys: loss, log_vars, num_samples.

  • loss is a tensor for back propagation, which can be a weighted sum of multiple losses.
  • log_vars contains all the variables to be sent to the logger.
  • num_samples indicates the batch size (when the model is DDP, it means the batch size on each GPU), which is used for averaging the logs.

Return type:

dict

val_step(data, optimizer=None)[source]

The iteration step during validation.

This method shares the same signature as train_step(), but used during val epochs. Note that the evaluation after training epochs is not implemented with this method, but an evaluation hook.

with_bbox

whether the detector has a bbox head

Type:bool
with_mask

whether the detector has a mask head

Type:bool
with_neck

whether the detector has a neck

Type:bool
with_shared_head

whether the detector has a shared head in the RoI Head

Type:bool
class mmdet.models.detectors.SingleStageDetector(backbone, neck=None, bbox_head=None, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Base class for single-stage detectors.

Single-stage detectors directly and densely predict bounding boxes on the output features of the backbone+neck.

aug_test(imgs, img_metas, rescale=False)[source]

Test function with test time augmentation.

Parameters:
  • imgs (list[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains all images in the batch.
  • img_metas (list[list[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch. each dict has image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

BBox results of each image and classes.

The outer list corresponds to each image. The inner list corresponds to each class.

Return type:

list[list[np.ndarray]]

extract_feat(img)[source]

Directly extract features from the backbone+neck.

forward_dummy(img)[source]

Used for computing network flops.

See mmdetection/tools/analysis_tools/get_flops.py

forward_train(img, img_metas, gt_bboxes, gt_labels, gt_bboxes_ignore=None)[source]
Parameters:
  • img (Tensor) – Input images of shape (N, C, H, W). Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – A List of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet.datasets.pipelines.Collect.
  • gt_bboxes (list[Tensor]) – Each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – Class indices corresponding to each box
  • gt_bboxes_ignore (None | list[Tensor]) – Specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

onnx_export(img, img_metas)[source]

Test function without test time augmentation.

Parameters:
  • img (torch.Tensor) – input images.
  • img_metas (list[dict]) – List of image information.
Returns:

dets of shape [N, num_det, 5]

and class labels of shape [N, num_det].

Return type:

tuple[Tensor, Tensor]

simple_test(img, img_metas, rescale=False)[source]

Test function without test-time augmentation.

Parameters:
  • img (torch.Tensor) – Images with shape (N, C, H, W).
  • img_metas (list[dict]) – List of image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

BBox results of each image and classes.

The outer list corresponds to each image. The inner list corresponds to each class.

Return type:

list[list[np.ndarray]]

class mmdet.models.detectors.TwoStageDetector(backbone, neck=None, rpn_head=None, roi_head=None, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Base class for two-stage detectors.

Two-stage detectors typically consisting of a region proposal network and a task-specific regression head.

async_simple_test(img, img_meta, proposals=None, rescale=False)[source]

Async test without augmentation.

aug_test(imgs, img_metas, rescale=False)[source]

Test with augmentations.

If rescale is False, then returned bboxes and masks will fit the scale of imgs[0].

extract_feat(img)[source]

Directly extract features from the backbone+neck.

forward_dummy(img)[source]

Used for computing network flops.

See mmdetection/tools/analysis_tools/get_flops.py

forward_train(img, img_metas, gt_bboxes, gt_labels, gt_bboxes_ignore=None, gt_masks=None, proposals=None, **kwargs)[source]
Parameters:
  • img (Tensor) – of shape (N, C, H, W) encoding input images. Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – list of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet/datasets/pipelines/formatting.py:Collect.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
  • gt_masks (None | Tensor) – true segmentation masks for each box used if the architecture supports a segmentation task.
  • proposals – override rpn proposals with custom proposals. Use when with_rpn is False.
Returns:

a dictionary of loss components

Return type:

dict[str, Tensor]

simple_test(img, img_metas, proposals=None, rescale=False)[source]

Test without augmentation.

with_roi_head

whether the detector has a RoI head

Type:bool
with_rpn

whether the detector has RPN

Type:bool
class mmdet.models.detectors.RPN(backbone, neck, rpn_head, train_cfg, test_cfg, pretrained=None, init_cfg=None)[source]

Implementation of Region Proposal Network.

aug_test(imgs, img_metas, rescale=False)[source]

Test function with test time augmentation.

Parameters:
  • imgs (list[torch.Tensor]) – List of multiple images
  • img_metas (list[dict]) – List of image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

proposals

Return type:

list[np.ndarray]

extract_feat(img)[source]

Extract features.

Parameters:img (torch.Tensor) – Image tensor with shape (n, c, h ,w).
Returns:
Multi-level features that may have
different resolutions.
Return type:list[torch.Tensor]
forward_dummy(img)[source]

Dummy forward function.

forward_train(img, img_metas, gt_bboxes=None, gt_bboxes_ignore=None)[source]
Parameters:
  • img (Tensor) – Input images of shape (N, C, H, W). Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – A List of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet.datasets.pipelines.Collect.
  • gt_bboxes (list[Tensor]) – Each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_bboxes_ignore (None | list[Tensor]) – Specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

show_result(data, result, top_k=20, **kwargs)[source]

Show RPN proposals on the image.

Parameters:
  • data (str or np.ndarray) – Image filename or loaded image.
  • result (Tensor or tuple) – The results to draw over img bbox_result or (bbox_result, segm_result).
  • top_k (int) – Plot the first k bboxes only if set positive. Default: 20
Returns:

The image with bboxes drawn on it.

Return type:

np.ndarray

simple_test(img, img_metas, rescale=False)[source]

Test function without test time augmentation.

Parameters:
  • imgs (list[torch.Tensor]) – List of multiple images
  • img_metas (list[dict]) – List of image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

proposals

Return type:

list[np.ndarray]

class mmdet.models.detectors.KnowledgeDistillationSingleStageDetector(backbone, neck, bbox_head, teacher_config, teacher_ckpt=None, eval_teacher=True, train_cfg=None, test_cfg=None, pretrained=None)[source]

Implementation of Distilling the Knowledge in a Neural Network..

Parameters:
  • teacher_config (str | dict) – Config file path or the config object of teacher model.
  • teacher_ckpt (str, optional) – Checkpoint path of teacher model. If left as None, the model will not load any weights.
cuda(device=None)[source]

Since teacher_model is registered as a plain object, it is necessary to put the teacher model to cuda when calling cuda function.

forward_train(img, img_metas, gt_bboxes, gt_labels, gt_bboxes_ignore=None)[source]
Parameters:
  • img (Tensor) – Input images of shape (N, C, H, W). Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – A List of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet.datasets.pipelines.Collect.
  • gt_bboxes (list[Tensor]) – Each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – Class indices corresponding to each box
  • gt_bboxes_ignore (None | list[Tensor]) – Specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

train(mode=True)[source]

Set the same train mode for teacher and student model.

class mmdet.models.detectors.FastRCNN(backbone, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Implementation of Fast R-CNN

forward_test(imgs, img_metas, proposals, **kwargs)[source]
Parameters:
  • imgs (List[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains all images in the batch.
  • img_metas (List[List[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch.
  • proposals (List[List[Tensor]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch. The Tensor should have a shape Px4, where P is the number of proposals.
class mmdet.models.detectors.FasterRCNN(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Implementation of Faster R-CNN

class mmdet.models.detectors.MaskRCNN(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Implementation of Mask R-CNN

class mmdet.models.detectors.CascadeRCNN(backbone, neck=None, rpn_head=None, roi_head=None, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of Cascade R-CNN: Delving into High Quality Object Detection

show_result(data, result, **kwargs)[source]

Show prediction results of the detector.

Parameters:
  • data (str or np.ndarray) – Image filename or loaded image.
  • result (Tensor or tuple) – The results to draw over img bbox_result or (bbox_result, segm_result).
Returns:

The image with bboxes drawn on it.

Return type:

np.ndarray

class mmdet.models.detectors.HybridTaskCascade(**kwargs)[source]

Implementation of HTC

with_semantic

whether the detector has a semantic head

Type:bool
class mmdet.models.detectors.RetinaNet(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of RetinaNet

class mmdet.models.detectors.FCOS(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of FCOS

class mmdet.models.detectors.GridRCNN(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Grid R-CNN.

This detector is the implementation of: - Grid R-CNN (https://arxiv.org/abs/1811.12030) - Grid R-CNN Plus: Faster and Better (https://arxiv.org/abs/1906.05688)

class mmdet.models.detectors.MaskScoringRCNN(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Mask Scoring RCNN.

https://arxiv.org/abs/1903.00241

class mmdet.models.detectors.RepPointsDetector(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

RepPoints: Point Set Representation for Object Detection.

This detector is the implementation of: - RepPoints detector (https://arxiv.org/pdf/1904.11490)

class mmdet.models.detectors.FOVEA(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of FoveaBox

class mmdet.models.detectors.FSAF(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of FSAF

class mmdet.models.detectors.NASFCOS(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

NAS-FCOS: Fast Neural Architecture Search for Object Detection.

https://arxiv.org/abs/1906.0442

class mmdet.models.detectors.PointRend(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

PointRend: Image Segmentation as Rendering

This detector is the implementation of PointRend.

class mmdet.models.detectors.GFL(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]
class mmdet.models.detectors.CornerNet(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

CornerNet.

This detector is the implementation of the paper CornerNet: Detecting Objects as Paired Keypoints .

aug_test(imgs, img_metas, rescale=False)[source]

Augment testing of CornerNet.

Parameters:
  • imgs (list[Tensor]) – Augmented images.
  • img_metas (list[list[dict]]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • rescale (bool) – If True, return boxes in original image space. Default: False.

Note

imgs must including flipped image pairs.

Returns:
BBox results of each image and classes.
The outer list corresponds to each image. The inner list corresponds to each class.
Return type:list[list[np.ndarray]]
merge_aug_results(aug_results, img_metas)[source]

Merge augmented detection bboxes and score.

Parameters:
  • aug_results (list[list[Tensor]]) – Det_bboxes and det_labels of each image.
  • img_metas (list[list[dict]]) – Meta information of each image, e.g., image size, scaling factor, etc.
Returns:

(bboxes, labels)

Return type:

tuple

class mmdet.models.detectors.PAA(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of PAA.

class mmdet.models.detectors.YOLOV3(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]
class mmdet.models.detectors.YOLACT(backbone, neck, bbox_head, segm_head, mask_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of YOLACT

aug_test(imgs, img_metas, rescale=False)[source]

Test with augmentations.

forward_dummy(img)[source]

Used for computing network flops.

See mmdetection/tools/analysis_tools/get_flops.py

forward_train(img, img_metas, gt_bboxes, gt_labels, gt_bboxes_ignore=None, gt_masks=None)[source]
Parameters:
  • img (Tensor) – of shape (N, C, H, W) encoding input images. Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – list of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet/datasets/pipelines/formatting.py:Collect.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
  • gt_masks (None | Tensor) – true segmentation masks for each box used if the architecture supports a segmentation task.
Returns:

a dictionary of loss components

Return type:

dict[str, Tensor]

simple_test(img, img_metas, rescale=False)[source]

Test function without test-time augmentation.

class mmdet.models.detectors.VFNet(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of `VarifocalNet (VFNet).<https://arxiv.org/abs/2008.13367>`_

class mmdet.models.detectors.DETR(backbone, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of DETR: End-to-End Object Detection with Transformers

forward_dummy(img)[source]

Used for computing network flops.

See mmdetection/tools/analysis_tools/get_flops.py

onnx_export(img, img_metas)[source]

Test function for exporting to ONNX, without test time augmentation.

Parameters:
  • img (torch.Tensor) – input images.
  • img_metas (list[dict]) – List of image information.
Returns:

dets of shape [N, num_det, 5]

and class labels of shape [N, num_det].

Return type:

tuple[Tensor, Tensor]

class mmdet.models.detectors.TridentFasterRCNN(backbone, rpn_head, roi_head, train_cfg, test_cfg, neck=None, pretrained=None, init_cfg=None)[source]

Implementation of TridentNet

aug_test(imgs, img_metas, rescale=False)[source]

Test with augmentations.

If rescale is False, then returned bboxes and masks will fit the scale of imgs[0].

forward_train(img, img_metas, gt_bboxes, gt_labels, **kwargs)[source]

make copies of img and gts to fit multi-branch.

simple_test(img, img_metas, proposals=None, rescale=False)[source]

Test without augmentation.

class mmdet.models.detectors.SparseRCNN(*args, **kwargs)[source]

Implementation of Sparse R-CNN: End-to-End Object Detection with Learnable Proposals

forward_dummy(img)[source]

Used for computing network flops.

See mmdetection/tools/analysis_tools/get_flops.py

forward_train(img, img_metas, gt_bboxes, gt_labels, gt_bboxes_ignore=None, gt_masks=None, proposals=None, **kwargs)[source]

Forward function of SparseR-CNN in train stage.

Parameters:
  • img (Tensor) – of shape (N, C, H, W) encoding input images. Typically these should be mean centered and std scaled.
  • img_metas (list[dict]) – list of image info dict where each dict has: ‘img_shape’, ‘scale_factor’, ‘flip’, and may also contain ‘filename’, ‘ori_shape’, ‘pad_shape’, and ‘img_norm_cfg’. For details on the values of these keys see mmdet.datasets.pipelines.Collect.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • gt_bboxes_ignore (None | list[Tensor) – specify which bounding boxes can be ignored when computing the loss.
  • gt_masks (List[Tensor], optional) – Segmentation masks for each box. But we don’t support it in this architecture.
  • proposals (List[Tensor], optional) – override rpn proposals with custom proposals. Use when with_rpn is False.
Returns:

a dictionary of loss components

Return type:

dict[str, Tensor]

simple_test(img, img_metas, rescale=False)[source]

Test function without test time augmentation.

Parameters:
  • imgs (list[torch.Tensor]) – List of multiple images
  • img_metas (list[dict]) – List of image information.
  • rescale (bool) – Whether to rescale the results. Defaults to False.
Returns:

BBox results of each image and classes.

The outer list corresponds to each image. The inner list corresponds to each class.

Return type:

list[list[np.ndarray]]

class mmdet.models.detectors.SCNet(**kwargs)[source]

Implementation of SCNet

class mmdet.models.detectors.DeformableDETR(*args, **kwargs)[source]
class mmdet.models.detectors.AutoAssign(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None)[source]

Implementation of AutoAssign: Differentiable Label Assignment for Dense Object Detection.

class mmdet.models.detectors.YOLOF(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None)[source]

Implementation of You Only Look One-level Feature

class mmdet.models.detectors.CenterNet(backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None, init_cfg=None)[source]

Implementation of CenterNet(Objects as Points)

<https://arxiv.org/abs/1904.07850>.

aug_test(imgs, img_metas, rescale=True)[source]

Augment testing of CenterNet. Aug test must have flipped image pair, and unlike CornerNet, it will perform an averaging operation on the feature map instead of detecting bbox.

Parameters:
  • imgs (list[Tensor]) – Augmented images.
  • img_metas (list[list[dict]]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • rescale (bool) – If True, return boxes in original image space. Default: True.

Note

imgs must including flipped image pairs.

Returns:
BBox results of each image and classes.
The outer list corresponds to each image. The inner list corresponds to each class.
Return type:list[list[np.ndarray]]
merge_aug_results(aug_results, with_nms)[source]

Merge augmented detection bboxes and score.

Parameters:
  • aug_results (list[list[Tensor]]) – Det_bboxes and det_labels of each image.
  • with_nms (bool) – If True, do nms before return boxes.
Returns:

(out_bboxes, out_labels)

Return type:

tuple

backbones

class mmdet.models.backbones.RegNet(arch, in_channels=3, stem_channels=32, base_channels=32, strides=(2, 2, 2, 2), dilations=(1, 1, 1, 1), out_indices=(0, 1, 2, 3), style='pytorch', deep_stem=False, avg_down=False, frozen_stages=-1, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, norm_eval=True, dcn=None, stage_with_dcn=(False, False, False, False), plugins=None, with_cp=False, zero_init_residual=True, pretrained=None, init_cfg=None)[source]

RegNet backbone.

More details can be found in paper .

Parameters:
  • arch (dict) –

    The parameter of RegNets.

    • w0 (int): initial width
    • wa (float): slope of width
    • wm (float): quantization parameter to quantize the width
    • depth (int): depth of the backbone
    • group_w (int): width of group
    • bot_mul (float): bottleneck ratio, i.e. expansion of bottleneck.
  • strides (Sequence[int]) – Strides of the first block of each stage.
  • base_channels (int) – Base channels after stem layer.
  • in_channels (int) – Number of input image channels. Default: 3.
  • dilations (Sequence[int]) – Dilation of each stage.
  • out_indices (Sequence[int]) – Output from which stages.
  • style (str) – pytorch or caffe. If set to “pytorch”, the stride-two layer is the 3x3 conv layer, otherwise the stride-two layer is the first 1x1 conv layer.
  • frozen_stages (int) – Stages to be frozen (all param fixed). -1 means not freezing any parameters.
  • norm_cfg (dict) – dictionary to construct and config norm layer.
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • zero_init_residual (bool) – whether to use zero init for last norm layer in resblocks to let them behave as identity.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import RegNet
>>> import torch
>>> self = RegNet(
        arch=dict(
            w0=88,
            wa=26.31,
            wm=2.25,
            group_w=48,
            depth=25,
            bot_mul=1.0))
>>> self.eval()
>>> inputs = torch.rand(1, 3, 32, 32)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
(1, 96, 8, 8)
(1, 192, 4, 4)
(1, 432, 2, 2)
(1, 1008, 1, 1)
adjust_width_group(widths, bottleneck_ratio, groups)[source]

Adjusts the compatibility of widths and groups.

Parameters:
  • widths (list[int]) – Width of each stage.
  • bottleneck_ratio (float) – Bottleneck ratio.
  • groups (int) – number of groups in each stage
Returns:

The adjusted widths and groups of each stage.

Return type:

tuple(list)

forward(x)[source]

Forward function.

generate_regnet(initial_width, width_slope, width_parameter, depth, divisor=8)[source]

Generates per block width from RegNet parameters.

Parameters:
  • initial_width ([int]) – Initial width of the backbone
  • width_slope ([float]) – Slope of the quantized linear function
  • width_parameter ([int]) – Parameter used to quantize the width.
  • depth ([int]) – Depth of the backbone.
  • divisor (int, optional) – The divisor of channels. Defaults to 8.
Returns:

return a list of widths of each stage and the number of stages

Return type:

list, int

get_stages_from_blocks(widths)[source]

Gets widths/stage_blocks of network at each stage.

Parameters:widths (list[int]) – Width in each stage.
Returns:width and depth of each stage
Return type:tuple(list)
static quantize_float(number, divisor)[source]

Converts a float to closest non-zero int divisible by divisor.

Parameters:
  • number (int) – Original number to be quantized.
  • divisor (int) – Divisor used to quantize the number.
Returns:

quantized number that is divisible by devisor.

Return type:

int

class mmdet.models.backbones.ResNet(depth, in_channels=3, stem_channels=None, base_channels=64, num_stages=4, strides=(1, 2, 2, 2), dilations=(1, 1, 1, 1), out_indices=(0, 1, 2, 3), style='pytorch', deep_stem=False, avg_down=False, frozen_stages=-1, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, norm_eval=True, dcn=None, stage_with_dcn=(False, False, False, False), plugins=None, with_cp=False, zero_init_residual=True, pretrained=None, init_cfg=None)[source]

ResNet backbone.

Parameters:
  • depth (int) – Depth of resnet, from {18, 34, 50, 101, 152}.
  • stem_channels (int | None) – Number of stem channels. If not specified, it will be the same as base_channels. Default: None.
  • base_channels (int) – Number of base channels of res layer. Default: 64.
  • in_channels (int) – Number of input image channels. Default: 3.
  • num_stages (int) – Resnet stages. Default: 4.
  • strides (Sequence[int]) – Strides of the first block of each stage.
  • dilations (Sequence[int]) – Dilation of each stage.
  • out_indices (Sequence[int]) – Output from which stages.
  • style (str) – pytorch or caffe. If set to “pytorch”, the stride-two layer is the 3x3 conv layer, otherwise the stride-two layer is the first 1x1 conv layer.
  • deep_stem (bool) – Replace 7x7 conv in input stem with 3 3x3 conv
  • avg_down (bool) – Use AvgPool instead of stride conv when downsampling in the bottleneck.
  • frozen_stages (int) – Stages to be frozen (stop grad and set eval mode). -1 means not freezing any parameters.
  • norm_cfg (dict) – Dictionary to construct and config norm layer.
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • plugins (list[dict]) –

    List of plugins for stages, each dict contains:

    • cfg (dict, required): Cfg dict to build plugin.
    • position (str, required): Position inside block to insert plugin, options are ‘after_conv1’, ‘after_conv2’, ‘after_conv3’.
    • stages (tuple[bool], optional): Stages to apply plugin, length should be same as ‘num_stages’.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • zero_init_residual (bool) – Whether to use zero init for last norm layer in resblocks to let them behave as identity.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import ResNet
>>> import torch
>>> self = ResNet(depth=18)
>>> self.eval()
>>> inputs = torch.rand(1, 3, 32, 32)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
(1, 64, 8, 8)
(1, 128, 4, 4)
(1, 256, 2, 2)
(1, 512, 1, 1)
forward(x)[source]

Forward function.

make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer.

make_stage_plugins(plugins, stage_idx)[source]

Make plugins for ResNet stage_idx th stage.

Currently we support to insert context_block, empirical_attention_block, nonlocal_block into the backbone like ResNet/ResNeXt. They could be inserted after conv1/conv2/conv3 of Bottleneck.

An example of plugins format could be:

Examples

>>> plugins=[
...     dict(cfg=dict(type='xxx', arg1='xxx'),
...          stages=(False, True, True, True),
...          position='after_conv2'),
...     dict(cfg=dict(type='yyy'),
...          stages=(True, True, True, True),
...          position='after_conv3'),
...     dict(cfg=dict(type='zzz', postfix='1'),
...          stages=(True, True, True, True),
...          position='after_conv3'),
...     dict(cfg=dict(type='zzz', postfix='2'),
...          stages=(True, True, True, True),
...          position='after_conv3')
... ]
>>> self = ResNet(depth=18)
>>> stage_plugins = self.make_stage_plugins(plugins, 0)
>>> assert len(stage_plugins) == 3

Suppose stage_idx=0, the structure of blocks in the stage would be:

conv1-> conv2->conv3->yyy->zzz1->zzz2

Suppose ‘stage_idx=1’, the structure of blocks in the stage would be:

conv1-> conv2->xxx->conv3->yyy->zzz1->zzz2

If stages is missing, the plugin would be applied to all stages.

Parameters:
  • plugins (list[dict]) – List of plugins cfg to build. The postfix is required if multiple same type plugins are inserted.
  • stage_idx (int) – Index of stage to build
Returns:

Plugins for current stage

Return type:

list[dict]

norm1

the normalization layer named “norm1”

Type:nn.Module
train(mode=True)[source]

Convert the model into training mode while keep normalization layer freezed.

class mmdet.models.backbones.ResNetV1d(**kwargs)[source]

ResNetV1d variant described in Bag of Tricks.

Compared with default ResNet(ResNetV1b), ResNetV1d replaces the 7x7 conv in the input stem with three 3x3 convs. And in the downsampling block, a 2x2 avg_pool with stride 2 is added before conv, whose stride is changed to 1.

class mmdet.models.backbones.ResNeXt(groups=1, base_width=4, **kwargs)[source]

ResNeXt backbone.

Parameters:
  • depth (int) – Depth of resnet, from {18, 34, 50, 101, 152}.
  • in_channels (int) – Number of input image channels. Default: 3.
  • num_stages (int) – Resnet stages. Default: 4.
  • groups (int) – Group of resnext.
  • base_width (int) – Base width of resnext.
  • strides (Sequence[int]) – Strides of the first block of each stage.
  • dilations (Sequence[int]) – Dilation of each stage.
  • out_indices (Sequence[int]) – Output from which stages.
  • style (str) – pytorch or caffe. If set to “pytorch”, the stride-two layer is the 3x3 conv layer, otherwise the stride-two layer is the first 1x1 conv layer.
  • frozen_stages (int) – Stages to be frozen (all param fixed). -1 means not freezing any parameters.
  • norm_cfg (dict) – dictionary to construct and config norm layer.
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • zero_init_residual (bool) – whether to use zero init for last norm layer in resblocks to let them behave as identity.
make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer

class mmdet.models.backbones.SSDVGG(depth, with_last_pool=False, ceil_mode=True, out_indices=(3, 4), out_feature_indices=(22, 34), pretrained=None, init_cfg=None, input_size=None, l2_norm_scale=None)[source]

VGG Backbone network for single-shot-detection.

Parameters:
  • depth (int) – Depth of vgg, from {11, 13, 16, 19}.
  • with_last_pool (bool) – Whether to add a pooling layer at the last of the model
  • ceil_mode (bool) – When True, will use ceil instead of floor to compute the output shape.
  • out_indices (Sequence[int]) – Output from which stages.
  • out_feature_indices (Sequence[int]) – Output from which feature map.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
  • input_size (int, optional) – Deprecated argumment. Width and height of input, from {300, 512}.
  • l2_norm_scale (float, optional) – Deprecated argumment. L2 normalization layer init scale.

Example

>>> self = SSDVGG(input_size=300, depth=11)
>>> self.eval()
>>> inputs = torch.rand(1, 3, 300, 300)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
(1, 1024, 19, 19)
(1, 512, 10, 10)
(1, 256, 5, 5)
(1, 256, 3, 3)
(1, 256, 1, 1)
forward(x)[source]

Forward function.

init_weights(pretrained=None)[source]

Initialize the weights.

class mmdet.models.backbones.HRNet(extra, in_channels=3, conv_cfg=None, norm_cfg={'type': 'BN'}, norm_eval=True, with_cp=False, zero_init_residual=False, pretrained=None, init_cfg=None)[source]

HRNet backbone.

High-Resolution Representations for Labeling Pixels and Regions arXiv: https://arxiv.org/abs/1904.04514

Parameters:
  • extra (dict) – detailed configuration for each stage of HRNet.
  • in_channels (int) – Number of input image channels. Default: 3.
  • conv_cfg (dict) – dictionary to construct and config conv layer.
  • norm_cfg (dict) – dictionary to construct and config norm layer.
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • zero_init_residual (bool) – whether to use zero init for last norm layer in resblocks to let them behave as identity.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import HRNet
>>> import torch
>>> extra = dict(
>>>     stage1=dict(
>>>         num_modules=1,
>>>         num_branches=1,
>>>         block='BOTTLENECK',
>>>         num_blocks=(4, ),
>>>         num_channels=(64, )),
>>>     stage2=dict(
>>>         num_modules=1,
>>>         num_branches=2,
>>>         block='BASIC',
>>>         num_blocks=(4, 4),
>>>         num_channels=(32, 64)),
>>>     stage3=dict(
>>>         num_modules=4,
>>>         num_branches=3,
>>>         block='BASIC',
>>>         num_blocks=(4, 4, 4),
>>>         num_channels=(32, 64, 128)),
>>>     stage4=dict(
>>>         num_modules=3,
>>>         num_branches=4,
>>>         block='BASIC',
>>>         num_blocks=(4, 4, 4, 4),
>>>         num_channels=(32, 64, 128, 256)))
>>> self = HRNet(extra, in_channels=1)
>>> self.eval()
>>> inputs = torch.rand(1, 1, 32, 32)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
(1, 32, 8, 8)
(1, 64, 4, 4)
(1, 128, 2, 2)
(1, 256, 1, 1)
forward(x)[source]

Forward function.

norm1

the normalization layer named “norm1”

Type:nn.Module
norm2

the normalization layer named “norm2”

Type:nn.Module
train(mode=True)[source]

Convert the model into training mode will keeping the normalization layer freezed.

class mmdet.models.backbones.MobileNetV2(widen_factor=1.0, out_indices=(1, 2, 4, 7), frozen_stages=-1, conv_cfg=None, norm_cfg={'type': 'BN'}, act_cfg={'type': 'ReLU6'}, norm_eval=False, with_cp=False, pretrained=None, init_cfg=None)[source]

MobileNetV2 backbone.

Parameters:
  • widen_factor (float) – Width multiplier, multiply number of channels in each layer by this amount. Default: 1.0.
  • out_indices (Sequence[int], optional) – Output from which stages. Default: (1, 2, 4, 7).
  • frozen_stages (int) – Stages to be frozen (all param fixed). Default: -1, which means not freezing any parameters.
  • conv_cfg (dict, optional) – Config dict for convolution layer. Default: None, which means using conv2d.
  • norm_cfg (dict) – Config dict for normalization layer. Default: dict(type=’BN’).
  • act_cfg (dict) – Config dict for activation layer. Default: dict(type=’ReLU6’).
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only. Default: False.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed. Default: False.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
forward(x)[source]

Forward function.

make_layer(out_channels, num_blocks, stride, expand_ratio)[source]

Stack InvertedResidual blocks to build a layer for MobileNetV2.

Parameters:
  • out_channels (int) – out_channels of block.
  • num_blocks (int) – number of blocks.
  • stride (int) – stride of the first block. Default: 1
  • expand_ratio (int) – Expand the number of channels of the hidden layer in InvertedResidual by this ratio. Default: 6.
train(mode=True)[source]

Convert the model into training mode while keep normalization layer frozen.

class mmdet.models.backbones.Res2Net(scales=4, base_width=26, style='pytorch', deep_stem=True, avg_down=True, pretrained=None, init_cfg=None, **kwargs)[source]

Res2Net backbone.

Parameters:
  • scales (int) – Scales used in Res2Net. Default: 4
  • base_width (int) – Basic width of each scale. Default: 26
  • depth (int) – Depth of res2net, from {50, 101, 152}.
  • in_channels (int) – Number of input image channels. Default: 3.
  • num_stages (int) – Res2net stages. Default: 4.
  • strides (Sequence[int]) – Strides of the first block of each stage.
  • dilations (Sequence[int]) – Dilation of each stage.
  • out_indices (Sequence[int]) – Output from which stages.
  • style (str) – pytorch or caffe. If set to “pytorch”, the stride-two layer is the 3x3 conv layer, otherwise the stride-two layer is the first 1x1 conv layer.
  • deep_stem (bool) – Replace 7x7 conv in input stem with 3 3x3 conv
  • avg_down (bool) – Use AvgPool instead of stride conv when downsampling in the bottle2neck.
  • frozen_stages (int) – Stages to be frozen (stop grad and set eval mode). -1 means not freezing any parameters.
  • norm_cfg (dict) – Dictionary to construct and config norm layer.
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • plugins (list[dict]) –

    List of plugins for stages, each dict contains:

    • cfg (dict, required): Cfg dict to build plugin.
    • position (str, required): Position inside block to insert plugin, options are ‘after_conv1’, ‘after_conv2’, ‘after_conv3’.
    • stages (tuple[bool], optional): Stages to apply plugin, length should be same as ‘num_stages’.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • zero_init_residual (bool) – Whether to use zero init for last norm layer in resblocks to let them behave as identity.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import Res2Net
>>> import torch
>>> self = Res2Net(depth=50, scales=4, base_width=26)
>>> self.eval()
>>> inputs = torch.rand(1, 3, 32, 32)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
(1, 256, 8, 8)
(1, 512, 4, 4)
(1, 1024, 2, 2)
(1, 2048, 1, 1)
make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer.

class mmdet.models.backbones.HourglassNet(downsample_times=5, num_stacks=2, stage_channels=(256, 256, 384, 384, 384, 512), stage_blocks=(2, 2, 2, 2, 2, 4), feat_channel=256, norm_cfg={'requires_grad': True, 'type': 'BN'}, pretrained=None, init_cfg=None)[source]

HourglassNet backbone.

Stacked Hourglass Networks for Human Pose Estimation. More details can be found in the paper .

Parameters:
  • downsample_times (int) – Downsample times in a HourglassModule.
  • num_stacks (int) – Number of HourglassModule modules stacked, 1 for Hourglass-52, 2 for Hourglass-104.
  • stage_channels (list[int]) – Feature channel of each sub-module in a HourglassModule.
  • stage_blocks (list[int]) – Number of sub-modules stacked in a HourglassModule.
  • feat_channel (int) – Feature channel of conv after a HourglassModule.
  • norm_cfg (dict) – Dictionary to construct and config norm layer.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import HourglassNet
>>> import torch
>>> self = HourglassNet()
>>> self.eval()
>>> inputs = torch.rand(1, 3, 511, 511)
>>> level_outputs = self.forward(inputs)
>>> for level_output in level_outputs:
...     print(tuple(level_output.shape))
(1, 256, 128, 128)
(1, 256, 128, 128)
forward(x)[source]

Forward function.

init_weights()[source]

Init module weights.

class mmdet.models.backbones.DetectoRS_ResNet(sac=None, stage_with_sac=(False, False, False, False), rfp_inplanes=None, output_img=False, pretrained=None, init_cfg=None, **kwargs)[source]

ResNet backbone for DetectoRS.

Parameters:
  • sac (dict, optional) – Dictionary to construct SAC (Switchable Atrous Convolution). Default: None.
  • stage_with_sac (list) – Which stage to use sac. Default: (False, False, False, False).
  • rfp_inplanes (int, optional) – The number of channels from RFP. Default: None. If specified, an additional conv layer will be added for rfp_feat. Otherwise, the structure is the same as base class.
  • output_img (bool) – If True, the input image will be inserted into the starting position of output. Default: False.
forward(x)[source]

Forward function.

init_weights()[source]

Initialize the weights.

make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer for DetectoRS.

rfp_forward(x, rfp_feats)[source]

Forward function for RFP.

class mmdet.models.backbones.DetectoRS_ResNeXt(groups=1, base_width=4, **kwargs)[source]

ResNeXt backbone for DetectoRS.

Parameters:
  • groups (int) – The number of groups in ResNeXt.
  • base_width (int) – The base width of ResNeXt.
make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer for DetectoRS.

class mmdet.models.backbones.Darknet(depth=53, out_indices=(3, 4, 5), frozen_stages=-1, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, act_cfg={'negative_slope': 0.1, 'type': 'LeakyReLU'}, norm_eval=True, pretrained=None, init_cfg=None)[source]

Darknet backbone.

Parameters:
  • depth (int) – Depth of Darknet. Currently only support 53.
  • out_indices (Sequence[int]) – Output from which stages.
  • frozen_stages (int) – Stages to be frozen (stop grad and set eval mode). -1 means not freezing any parameters. Default: -1.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Dictionary to construct and config norm layer. Default: dict(type=’BN’, requires_grad=True)
  • act_cfg (dict) – Config dict for activation layer. Default: dict(type=’LeakyReLU’, negative_slope=0.1).
  • norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only.
  • pretrained (str, optional) – model pretrained path. Default: None
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None

Example

>>> from mmdet.models import Darknet
>>> import torch
>>> self = Darknet(depth=53)
>>> self.eval()
>>> inputs = torch.rand(1, 3, 416, 416)
>>> level_outputs = self.forward(inputs)
>>> for level_out in level_outputs:
...     print(tuple(level_out.shape))
...
(1, 256, 52, 52)
(1, 512, 26, 26)
(1, 1024, 13, 13)
forward(x)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

static make_conv_res_block(in_channels, out_channels, res_repeat, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, act_cfg={'negative_slope': 0.1, 'type': 'LeakyReLU'})[source]

In Darknet backbone, ConvLayer is usually followed by ResBlock. This function will make that. The Conv layers always have 3x3 filters with stride=2. The number of the filters in Conv layer is the same as the out channels of the ResBlock.

Parameters:
  • in_channels (int) – The number of input channels.
  • out_channels (int) – The number of output channels.
  • res_repeat (int) – The number of ResBlocks.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Dictionary to construct and config norm layer. Default: dict(type=’BN’, requires_grad=True)
  • act_cfg (dict) – Config dict for activation layer. Default: dict(type=’LeakyReLU’, negative_slope=0.1).
train(mode=True)[source]

Sets the module in training mode.

This has any effect only on certain modules. See documentations of particular modules for details of their behaviors in training/evaluation mode, if they are affected, e.g. Dropout, BatchNorm, etc.

Parameters:mode (bool) – whether to set training mode (True) or evaluation mode (False). Default: True.
Returns:self
Return type:Module
class mmdet.models.backbones.ResNeSt(groups=1, base_width=4, radix=2, reduction_factor=4, avg_down_stride=True, **kwargs)[source]

ResNeSt backbone.

Parameters:
  • groups (int) – Number of groups of Bottleneck. Default: 1
  • base_width (int) – Base width of Bottleneck. Default: 4
  • radix (int) – Radix of SplitAttentionConv2d. Default: 2
  • reduction_factor (int) – Reduction factor of inter_channels in SplitAttentionConv2d. Default: 4.
  • avg_down_stride (bool) – Whether to use average pool for stride in Bottleneck. Default: True.
  • kwargs (dict) – Keyword arguments for ResNet.
make_res_layer(**kwargs)[source]

Pack all blocks in a stage into a ResLayer.

class mmdet.models.backbones.TridentResNet(depth, num_branch, test_branch_idx, trident_dilations, **kwargs)[source]

The stem layer, stage 1 and stage 2 in Trident ResNet are identical to ResNet, while in stage 3, Trident BottleBlock is utilized to replace the normal BottleBlock to yield trident output. Different branch shares the convolution weight but uses different dilations to achieve multi-scale output.

/ stage3(b0) x - stem - stage1 - stage2 - stage3(b1) - output stage3(b2) /
Parameters:
  • depth (int) – Depth of resnet, from {50, 101, 152}.
  • num_branch (int) – Number of branches in TridentNet.
  • test_branch_idx (int) – In inference, all 3 branches will be used if test_branch_idx==-1, otherwise only branch with index test_branch_idx will be used.
  • trident_dilations (tuple[int]) – Dilations of different trident branch. len(trident_dilations) should be equal to num_branch.

necks

class mmdet.models.necks.FPN(in_channels, out_channels, num_outs, start_level=0, end_level=-1, add_extra_convs=False, relu_before_extra_convs=False, no_norm_on_lateral=False, conv_cfg=None, norm_cfg=None, act_cfg=None, upsample_cfg={'mode': 'nearest'}, init_cfg={'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'})[source]

Feature Pyramid Network.

This is an implementation of paper Feature Pyramid Networks for Object Detection.

Parameters:
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (int) – Number of output channels (used at each scale)
  • num_outs (int) – Number of output scales.
  • start_level (int) – Index of the start input backbone level used to build the feature pyramid. Default: 0.
  • end_level (int) – Index of the end input backbone level (exclusive) to build the feature pyramid. Default: -1, which means the last level.
  • add_extra_convs (bool | str) –

    If bool, it decides whether to add conv layers on top of the original feature maps. Default to False. If True, it is equivalent to add_extra_convs=’on_input’. If str, it specifies the source feature map of the extra convs. Only the following options are allowed

    • ’on_input’: Last feat map of neck inputs (i.e. backbone feature).
    • ’on_lateral’: Last feature map after lateral convs.
    • ’on_output’: The last output feature map after fpn convs.
  • relu_before_extra_convs (bool) – Whether to apply relu before the extra conv. Default: False.
  • no_norm_on_lateral (bool) – Whether to apply norm on lateral. Default: False.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Config dict for normalization layer. Default: None.
  • act_cfg (str) – Config dict for activation layer in ConvModule. Default: None.
  • upsample_cfg (dict) – Config dict for interpolate layer. Default: dict(mode=’nearest’)
  • init_cfg (dict or list[dict], optional) – Initialization config dict.

Example

>>> import torch
>>> in_channels = [2, 3, 5, 7]
>>> scales = [340, 170, 84, 43]
>>> inputs = [torch.rand(1, c, s, s)
...           for c, s in zip(in_channels, scales)]
>>> self = FPN(in_channels, 11, len(in_channels)).eval()
>>> outputs = self.forward(inputs)
>>> for i in range(len(outputs)):
...     print(f'outputs[{i}].shape = {outputs[i].shape}')
outputs[0].shape = torch.Size([1, 11, 340, 340])
outputs[1].shape = torch.Size([1, 11, 170, 170])
outputs[2].shape = torch.Size([1, 11, 84, 84])
outputs[3].shape = torch.Size([1, 11, 43, 43])
forward(inputs)[source]

Forward function.

class mmdet.models.necks.BFP(Balanced Feature Pyramids)[source]

BFP takes multi-level features as inputs and gather them into a single one, then refine the gathered feature and scatter the refined results to multi-level features. This module is used in Libra R-CNN (CVPR 2019), see the paper Libra R-CNN: Towards Balanced Learning for Object Detection for details.

Parameters:
  • in_channels (int) – Number of input channels (feature maps of all levels should have the same channels).
  • num_levels (int) – Number of input feature levels.
  • conv_cfg (dict) – The config dict for convolution layers.
  • norm_cfg (dict) – The config dict for normalization layers.
  • refine_level (int) – Index of integration and refine level of BSF in multi-level features from bottom to top.
  • refine_type (str) – Type of the refine op, currently support [None, ‘conv’, ‘non_local’].
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Forward function.

class mmdet.models.necks.ChannelMapper(in_channels, out_channels, kernel_size=3, conv_cfg=None, norm_cfg=None, act_cfg={'type': 'ReLU'}, num_outs=None, init_cfg={'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'})[source]

Channel Mapper to reduce/increase channels of backbone features.

This is used to reduce/increase channels of backbone features.

Parameters:
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (int) – Number of output channels (used at each scale).
  • kernel_size (int, optional) – kernel_size for reducing channels (used at each scale). Default: 3.
  • conv_cfg (dict, optional) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict, optional) – Config dict for normalization layer. Default: None.
  • act_cfg (dict, optional) – Config dict for activation layer in ConvModule. Default: dict(type=’ReLU’).
  • num_outs (int, optional) – Number of output feature maps. There would be extra_convs when num_outs larger than the length of in_channels.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.

Example

>>> import torch
>>> in_channels = [2, 3, 5, 7]
>>> scales = [340, 170, 84, 43]
>>> inputs = [torch.rand(1, c, s, s)
...           for c, s in zip(in_channels, scales)]
>>> self = ChannelMapper(in_channels, 11, 3).eval()
>>> outputs = self.forward(inputs)
>>> for i in range(len(outputs)):
...     print(f'outputs[{i}].shape = {outputs[i].shape}')
outputs[0].shape = torch.Size([1, 11, 340, 340])
outputs[1].shape = torch.Size([1, 11, 170, 170])
outputs[2].shape = torch.Size([1, 11, 84, 84])
outputs[3].shape = torch.Size([1, 11, 43, 43])
forward(inputs)[source]

Forward function.

class mmdet.models.necks.HRFPN(High Resolution Feature Pyramids)[source]

paper: High-Resolution Representations for Labeling Pixels and Regions.

Parameters:
  • in_channels (list) – number of channels for each branch.
  • out_channels (int) – output channels of feature pyramids.
  • num_outs (int) – number of output stages.
  • pooling_type (str) – pooling for generating feature pyramids from {MAX, AVG}.
  • conv_cfg (dict) – dictionary to construct and config conv layer.
  • norm_cfg (dict) – dictionary to construct and config norm layer.
  • with_cp (bool) – Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed.
  • stride (int) – stride of 3x3 convolutional layers
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Forward function.

class mmdet.models.necks.NASFPN(in_channels, out_channels, num_outs, stack_times, start_level=0, end_level=-1, add_extra_convs=False, norm_cfg=None, init_cfg={'layer': 'Conv2d', 'type': 'Caffe2Xavier'})[source]

NAS-FPN.

Implementation of NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection

Parameters:
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (int) – Number of output channels (used at each scale)
  • num_outs (int) – Number of output scales.
  • stack_times (int) – The number of times the pyramid architecture will be stacked.
  • start_level (int) – Index of the start input backbone level used to build the feature pyramid. Default: 0.
  • end_level (int) – Index of the end input backbone level (exclusive) to build the feature pyramid. Default: -1, which means the last level.
  • add_extra_convs (bool) – It decides whether to add conv layers on top of the original feature maps. Default to False. If True, its actual mode is specified by extra_convs_on_inputs.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Forward function.

class mmdet.models.necks.FPN_CARAFE(in_channels, out_channels, num_outs, start_level=0, end_level=-1, norm_cfg=None, act_cfg=None, order=('conv', 'norm', 'act'), upsample_cfg={'encoder_dilation': 1, 'encoder_kernel': 3, 'type': 'carafe', 'up_group': 1, 'up_kernel': 5}, init_cfg=None)[source]

FPN_CARAFE is a more flexible implementation of FPN. It allows more choice for upsample methods during the top-down pathway.

It can reproduce the performance of ICCV 2019 paper CARAFE: Content-Aware ReAssembly of FEatures Please refer to https://arxiv.org/abs/1905.02188 for more details.

Parameters:
  • in_channels (list[int]) – Number of channels for each input feature map.
  • out_channels (int) – Output channels of feature pyramids.
  • num_outs (int) – Number of output stages.
  • start_level (int) – Start level of feature pyramids. (Default: 0)
  • end_level (int) – End level of feature pyramids. (Default: -1 indicates the last level).
  • norm_cfg (dict) – Dictionary to construct and config norm layer.
  • activate (str) – Type of activation function in ConvModule (Default: None indicates w/o activation).
  • order (dict) – Order of components in ConvModule.
  • upsample (str) – Type of upsample layer.
  • upsample_cfg (dict) – Dictionary to construct and config upsample layer.
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
forward(inputs)[source]

Forward function.

init_weights()[source]

Initialize the weights of module.

slice_as(src, dst)[source]

Slice src as dst

Note

src should have the same or larger size than dst.

Parameters:
  • src (torch.Tensor) – Tensors to be sliced.
  • dst (torch.Tensor) – src will be sliced to have the same size as dst.
Returns:

Sliced tensor.

Return type:

torch.Tensor

tensor_add(a, b)[source]

Add tensors a and b that might have different sizes.

class mmdet.models.necks.PAFPN(in_channels, out_channels, num_outs, start_level=0, end_level=-1, add_extra_convs=False, relu_before_extra_convs=False, no_norm_on_lateral=False, conv_cfg=None, norm_cfg=None, act_cfg=None, init_cfg={'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'})[source]

Path Aggregation Network for Instance Segmentation.

This is an implementation of the PAFPN in Path Aggregation Network.

Parameters:
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (int) – Number of output channels (used at each scale)
  • num_outs (int) – Number of output scales.
  • start_level (int) – Index of the start input backbone level used to build the feature pyramid. Default: 0.
  • end_level (int) – Index of the end input backbone level (exclusive) to build the feature pyramid. Default: -1, which means the last level.
  • add_extra_convs (bool | str) –

    If bool, it decides whether to add conv layers on top of the original feature maps. Default to False. If True, it is equivalent to add_extra_convs=’on_input’. If str, it specifies the source feature map of the extra convs. Only the following options are allowed

    • ’on_input’: Last feat map of neck inputs (i.e. backbone feature).
    • ’on_lateral’: Last feature map after lateral convs.
    • ’on_output’: The last output feature map after fpn convs.
  • relu_before_extra_convs (bool) – Whether to apply relu before the extra conv. Default: False.
  • no_norm_on_lateral (bool) – Whether to apply norm on lateral. Default: False.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Config dict for normalization layer. Default: None.
  • act_cfg (str) – Config dict for activation layer in ConvModule. Default: None.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Forward function.

class mmdet.models.necks.NASFCOS_FPN(in_channels, out_channels, num_outs, start_level=1, end_level=-1, add_extra_convs=False, conv_cfg=None, norm_cfg=None, init_cfg=None)[source]

FPN structure in NASFPN.

Implementation of paper NAS-FCOS: Fast Neural Architecture Search for Object Detection

Parameters:
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (int) – Number of output channels (used at each scale)
  • num_outs (int) – Number of output scales.
  • start_level (int) – Index of the start input backbone level used to build the feature pyramid. Default: 0.
  • end_level (int) – Index of the end input backbone level (exclusive) to build the feature pyramid. Default: -1, which means the last level.
  • add_extra_convs (bool) – It decides whether to add conv layers on top of the original feature maps. Default to False. If True, its actual mode is specified by extra_convs_on_inputs.
  • conv_cfg (dict) – dictionary to construct and config conv layer.
  • norm_cfg (dict) – dictionary to construct and config norm layer.
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
forward(inputs)[source]

Forward function.

init_weights()[source]

Initialize the weights of module.

class mmdet.models.necks.RFP(Recursive Feature Pyramid)[source]

This is an implementation of RFP in DetectoRS. Different from standard FPN, the input of RFP should be multi level features along with origin input image of backbone.

Parameters:
  • rfp_steps (int) – Number of unrolled steps of RFP.
  • rfp_backbone (dict) – Configuration of the backbone for RFP.
  • aspp_out_channels (int) – Number of output channels of ASPP module.
  • aspp_dilations (tuple[int]) – Dilation rates of four branches. Default: (1, 3, 6, 1)
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
forward(inputs)[source]

Forward function.

init_weights()[source]

Initialize the weights.

class mmdet.models.necks.YOLOV3Neck(num_scales, in_channels, out_channels, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, act_cfg={'negative_slope': 0.1, 'type': 'LeakyReLU'}, init_cfg=None)[source]

The neck of YOLOV3.

It can be treated as a simplified version of FPN. It will take the result from Darknet backbone and do some upsampling and concatenation. It will finally output the detection result.

Note

The input feats should be from top to bottom.
i.e., from high-lvl to low-lvl
But YOLOV3Neck will process them in reversed order.
i.e., from bottom (high-lvl) to top (low-lvl)
Parameters:
  • num_scales (int) – The number of scales / stages.
  • in_channels (List[int]) – The number of input channels per scale.
  • out_channels (List[int]) – The number of output channels per scale.
  • conv_cfg (dict, optional) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict, optional) – Dictionary to construct and config norm layer. Default: dict(type=’BN’, requires_grad=True)
  • act_cfg (dict, optional) – Config dict for activation layer. Default: dict(type=’LeakyReLU’, negative_slope=0.1).
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
forward(feats)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mmdet.models.necks.FPG(in_channels, out_channels, num_outs, stack_times, paths, inter_channels=None, same_down_trans=None, same_up_trans={'kernel_size': 3, 'padding': 1, 'stride': 2, 'type': 'conv'}, across_lateral_trans={'kernel_size': 1, 'type': 'conv'}, across_down_trans={'kernel_size': 3, 'type': 'conv'}, across_up_trans=None, across_skip_trans={'type': 'identity'}, output_trans={'kernel_size': 3, 'type': 'last_conv'}, start_level=0, end_level=-1, add_extra_convs=False, norm_cfg=None, skip_inds=None, init_cfg=[{'type': 'Caffe2Xavier', 'layer': 'Conv2d'}, {'type': 'Constant', 'layer': ['_BatchNorm', '_InstanceNorm', 'GroupNorm', 'LayerNorm'], 'val': 1.0}])[source]

FPG.

Implementation of Feature Pyramid Grids (FPG). This implementation only gives the basic structure stated in the paper. But users can implement different type of transitions to fully explore the the potential power of the structure of FPG.

Parameters:
  • in_channels (int) – Number of input channels (feature maps of all levels should have the same channels).
  • out_channels (int) – Number of output channels (used at each scale)
  • num_outs (int) – Number of output scales.
  • stack_times (int) – The number of times the pyramid architecture will be stacked.
  • paths (list[str]) – Specify the path order of each stack level. Each element in the list should be either ‘bu’ (bottom-up) or ‘td’ (top-down).
  • inter_channels (int) – Number of inter channels.
  • same_up_trans (dict) – Transition that goes down at the same stage.
  • same_down_trans (dict) – Transition that goes up at the same stage.
  • across_lateral_trans (dict) – Across-pathway same-stage
  • across_down_trans (dict) – Across-pathway bottom-up connection.
  • across_up_trans (dict) – Across-pathway top-down connection.
  • across_skip_trans (dict) – Across-pathway skip connection.
  • output_trans (dict) – Transition that trans the output of the last stage.
  • start_level (int) – Index of the start input backbone level used to build the feature pyramid. Default: 0.
  • end_level (int) – Index of the end input backbone level (exclusive) to build the feature pyramid. Default: -1, which means the last level.
  • add_extra_convs (bool) – It decides whether to add conv layers on top of the original feature maps. Default to False. If True, its actual mode is specified by extra_convs_on_inputs.
  • norm_cfg (dict) – Config dict for normalization layer. Default: None.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mmdet.models.necks.DilatedEncoder(in_channels, out_channels, block_mid_channels, num_residual_blocks)[source]

Dilated Encoder for YOLOF <https://arxiv.org/abs/2103.09460>`.

This module contains two types of components:
  • the original FPN lateral convolution layer and fpn convolution layer,
    which are 1x1 conv + 3x3 conv
  • the dilated residual block
Parameters:
  • in_channels (int) – The number of input channels.
  • out_channels (int) – The number of output channels.
  • block_mid_channels (int) – The number of middle block output channels
  • num_residual_blocks (int) – The number of residual blocks.
forward(feature)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mmdet.models.necks.CTResNetNeck(in_channel, num_deconv_filters, num_deconv_kernels, use_dcn=True, init_cfg=None)[source]

The neck used in CenterNet for object classification and box regression.

Parameters:
  • in_channel (int) – Number of input channels.
  • num_deconv_filters (tuple[int]) – Number of filters per stage.
  • num_deconv_kernels (tuple[int]) – Number of kernels per stage.
  • use_dcn (bool) – If True, use DCNv2. Default: True.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

init_weights()[source]

Initialize the weights.

class mmdet.models.necks.SSDNeck(in_channels, out_channels, level_strides, level_paddings, l2_norm_scale=20.0, last_kernel_size=3, use_depthwise=False, conv_cfg=None, norm_cfg=None, act_cfg={'type': 'ReLU'}, init_cfg=[{'type': 'Xavier', 'distribution': 'uniform', 'layer': 'Conv2d'}, {'type': 'Constant', 'val': 1, 'layer': 'BatchNorm2d'}])[source]

Extra layers of SSD backbone to generate multi-scale feature maps.

Parameters:
  • in_channels (Sequence[int]) – Number of input channels per scale.
  • out_channels (Sequence[int]) – Number of output channels per scale.
  • level_strides (Sequence[int]) – Stride of 3x3 conv per level.
  • level_paddings (Sequence[int]) – Padding size of 3x3 conv per level.
  • l2_norm_scale (float|None) – L2 normalization layer init scale. If None, not use L2 normalization on the first input feature.
  • last_kernel_size (int) – Kernel size of the last conv layer. Default: 3.
  • use_depthwise (bool) – Whether to use DepthwiseSeparableConv. Default: False.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Dictionary to construct and config norm layer. Default: None.
  • act_cfg (dict) – Config dict for activation layer. Default: dict(type=’ReLU’).
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(inputs)[source]

Forward function.

dense_heads

class mmdet.models.dense_heads.AnchorFreeHead(num_classes, in_channels, feat_channels=256, stacked_convs=4, strides=(4, 8, 16, 32, 64), dcn_on_last_conv=False, conv_bias='auto', loss_cls={'alpha': 0.25, 'gamma': 2.0, 'loss_weight': 1.0, 'type': 'FocalLoss', 'use_sigmoid': True}, loss_bbox={'loss_weight': 1.0, 'type': 'IoULoss'}, conv_cfg=None, norm_cfg=None, train_cfg=None, test_cfg=None, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'conv_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'})[source]

Anchor-free head (FCOS, Fovea, RepPoints, etc.).

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • feat_channels (int) – Number of hidden channels. Used in child classes.
  • stacked_convs (int) – Number of stacking convs of the head.
  • strides (tuple) – Downsample factor of each feature map.
  • dcn_on_last_conv (bool) – If true, use dcn in the last layer of towers. Default: False.
  • conv_bias (bool | str) – If specified as auto, it will be decided by the norm_cfg. Bias of conv will be set as True if norm_cfg is None, otherwise False. Default: “auto”.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox (dict) – Config of localization loss.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Config dict for normalization layer. Default: None.
  • train_cfg (dict) – Training config of anchor head.
  • test_cfg (dict) – Testing config of anchor head.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
aug_test(feats, img_metas, rescale=False)[source]

Test function with test time augmentation.

Parameters:
  • feats (list[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains features for all images in the batch.
  • img_metas (list[list[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch. each dict has image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

bbox results of each class

Return type:

list[ndarray]

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
Usually contain classification scores and bbox predictions.
cls_scores (list[Tensor]): Box scores for each scale level,
each is a 4D-tensor, the channel number is num_points * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for each scale
level, each is a 4D-tensor, the channel number is num_points * 4.
Return type:tuple
forward_single(x)[source]

Forward features of a single scale level.

Parameters:x (Tensor) – FPN feature maps of the specified stride.
Returns:
Scores for each class, bbox predictions, features
after classification and regression conv layers, some models needs these features like FCOS.
Return type:tuple
get_bboxes(cls_scores, bbox_preds, img_metas, cfg=None, rescale=None)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_points * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_points * 4, H, W)
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space
get_points(featmap_sizes, dtype, device, flatten=False)[source]

Get points according to feature map sizes.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • dtype (torch.dtype) – Type of points.
  • device (torch.device) – Device of points.
Returns:

points of each image.

Return type:

tuple

get_targets(points, gt_bboxes_list, gt_labels_list)[source]

Compute regression, classification and centerness targets for points in multiple images.

Parameters:
  • points (list[Tensor]) – Points of each fpn level, each has shape (num_points, 2).
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image, each has shape (num_gt, 4).
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box, each has shape (num_gt,).
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level, each is a 4D-tensor, the channel number is num_points * num_classes.
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level, each is a 4D-tensor, the channel number is num_points * 4.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
class mmdet.models.dense_heads.AnchorHead(num_classes, in_channels, feat_channels=256, anchor_generator={'ratios': [0.5, 1.0, 2.0], 'scales': [8, 16, 32], 'strides': [4, 8, 16, 32, 64], 'type': 'AnchorGenerator'}, bbox_coder={'clip_border': True, 'target_means': (0.0, 0.0, 0.0, 0.0), 'target_stds': (1.0, 1.0, 1.0, 1.0), 'type': 'DeltaXYWHBBoxCoder'}, reg_decoded_bbox=False, loss_cls={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_bbox={'beta': 0.1111111111111111, 'loss_weight': 1.0, 'type': 'SmoothL1Loss'}, train_cfg=None, test_cfg=None, init_cfg={'layer': 'Conv2d', 'std': 0.01, 'type': 'Normal'})[source]

Anchor-based head (RPN, RetinaNet, SSD, etc.).

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • feat_channels (int) – Number of hidden channels. Used in child classes.
  • anchor_generator (dict) – Config dict for anchor generator
  • bbox_coder (dict) – Config of bounding box coder.
  • reg_decoded_bbox (bool) – If true, the regression loss would be applied directly on decoded bounding boxes, converting both the predicted boxes and regression targets to absolute coordinates format. Default False. It should be True when using IoULoss, GIoULoss, or DIoULoss in the bbox head.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox (dict) – Config of localization loss.
  • train_cfg (dict) – Training config of anchor head.
  • test_cfg (dict) – Testing config of anchor head.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
aug_test(feats, img_metas, rescale=False)[source]

Test function with test time augmentation.

Parameters:
  • feats (list[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains features for all images in the batch.
  • img_metas (list[list[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch. each dict has image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

Each item in result_list is 2-tuple.

The first item is bboxes with shape (n, 5), where 5 represent (tl_x, tl_y, br_x, br_y, score). The shape of the second tensor in the tuple is labels with shape (n,), The length of list should always be 1.

Return type:

list[tuple[Tensor, Tensor]]

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:A tuple of classification scores and bbox prediction.
  • cls_scores (list[Tensor]): Classification scores for all scale levels, each is a 4D-tensor, the channels number is num_anchors * num_classes.
  • bbox_preds (list[Tensor]): Box energies / deltas for all scale levels, each is a 4D-tensor, the channels number is num_anchors * 4.
Return type:tuple
forward_single(x)[source]

Forward feature of a single scale level.

Parameters:x (Tensor) – Features of a single scale level.
Returns:cls_score (Tensor): Cls scores for a single scale level the channels number is num_anchors * num_classes. bbox_pred (Tensor): Box energies / deltas for a single scale level, the channels number is num_anchors * 4.
Return type:tuple
get_anchors(featmap_sizes, img_metas, device='cuda')[source]

Get anchors according to feature map sizes.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • img_metas (list[dict]) – Image meta info.
  • device (torch.device | str) – Device for returned tensors
Returns:

anchor_list (list[Tensor]): Anchors of each image. valid_flag_list (list[Tensor]): Valid flags of each image.

Return type:

tuple

get_bboxes(cls_scores, bbox_preds, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each level in the feature pyramid, has shape (N, num_anchors * num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box energies / deltas for each level in the feature pyramid, has shape (N, num_anchors * 4, H, W).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1. The shape of the second tensor in the tuple is (n,), and each element represents the class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

Example

>>> import mmcv
>>> self = AnchorHead(
>>>     num_classes=9,
>>>     in_channels=1,
>>>     anchor_generator=dict(
>>>         type='AnchorGenerator',
>>>         scales=[8],
>>>         ratios=[0.5, 1.0, 2.0],
>>>         strides=[4,]))
>>> img_metas = [{'img_shape': (32, 32, 3), 'scale_factor': 1}]
>>> cfg = mmcv.Config(dict(
>>>     score_thr=0.00,
>>>     nms=dict(type='nms', iou_thr=1.0),
>>>     max_per_img=10))
>>> feat = torch.rand(1, 1, 3, 3)
>>> cls_score, bbox_pred = self.forward_single(feat)
>>> # note the input lists are over different levels, not images
>>> cls_scores, bbox_preds = [cls_score], [bbox_pred]
>>> result_list = self.get_bboxes(cls_scores, bbox_preds,
>>>                               img_metas, cfg)
>>> det_bboxes, det_labels = result_list[0]
>>> assert len(result_list) == 1
>>> assert det_bboxes.shape[1] == 5
>>> assert len(det_bboxes) == len(det_labels) == cfg.max_per_img
get_targets(anchor_list, valid_flag_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, label_channels=1, unmap_outputs=True, return_sampling_results=False)[source]

Compute regression and classification targets for anchors in multiple images.

Parameters:
  • anchor_list (list[list[Tensor]]) – Multi level anchors of each image. The outer list indicates images, and the inner list corresponds to feature levels of the image. Each element of the inner list is a tensor of shape (num_anchors, 4).
  • valid_flag_list (list[list[Tensor]]) – Multi level valid flags of each image. The outer list indicates images, and the inner list corresponds to feature levels of the image. Each element of the inner list is a tensor of shape (num_anchors, )
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • img_metas (list[dict]) – Meta info of each image.
  • gt_bboxes_ignore_list (list[Tensor]) – Ground truth bboxes to be ignored.
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box.
  • label_channels (int) – Channel of label.
  • unmap_outputs (bool) – Whether to map outputs back to the original set of anchors.
Returns:

Usually returns a tuple containing learning targets.

  • labels_list (list[Tensor]): Labels of each level.
  • label_weights_list (list[Tensor]): Label weights of each level.
  • bbox_targets_list (list[Tensor]): BBox targets of each level.
  • bbox_weights_list (list[Tensor]): BBox weights of each level.
  • num_total_pos (int): Number of positive samples in all images.
  • num_total_neg (int): Number of negative samples in all images.
additional_returns: This function enables user-defined returns from

self._get_targets_single. These returns are currently refined to properties at each feature map (i.e. having HxW dimension). The results will be concatenated after the end

Return type:

tuple

loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss. Default: None
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

loss_single(cls_score, bbox_pred, anchors, labels, label_weights, bbox_targets, bbox_weights, num_total_samples)[source]

Compute loss of a single scale level.

Parameters:
  • cls_score (Tensor) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W).
  • bbox_pred (Tensor) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W).
  • anchors (Tensor) – Box reference for each scale level with shape (N, num_total_anchors, 4).
  • labels (Tensor) – Labels of each anchors with shape (N, num_total_anchors).
  • label_weights (Tensor) – Label weights of each anchor with shape (N, num_total_anchors)
  • bbox_targets (Tensor) – BBox regression targets of each anchor wight shape (N, num_total_anchors, 4).
  • bbox_weights (Tensor) – BBox regression loss weights of each anchor with shape (N, num_total_anchors, 4).
  • num_total_samples (int) – If sampling, num total samples equal to the number of total anchors; Otherwise, it is the number of positive anchors.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.GuidedAnchorHead(num_classes, in_channels, feat_channels=256, approx_anchor_generator={'octave_base_scale': 8, 'ratios': [0.5, 1.0, 2.0], 'scales_per_octave': 3, 'strides': [4, 8, 16, 32, 64], 'type': 'AnchorGenerator'}, square_anchor_generator={'ratios': [1.0], 'scales': [8], 'strides': [4, 8, 16, 32, 64], 'type': 'AnchorGenerator'}, anchor_coder={'target_means': [0.0, 0.0, 0.0, 0.0], 'target_stds': [1.0, 1.0, 1.0, 1.0], 'type': 'DeltaXYWHBBoxCoder'}, bbox_coder={'target_means': [0.0, 0.0, 0.0, 0.0], 'target_stds': [1.0, 1.0, 1.0, 1.0], 'type': 'DeltaXYWHBBoxCoder'}, reg_decoded_bbox=False, deform_groups=4, loc_filter_thr=0.01, train_cfg=None, test_cfg=None, loss_loc={'alpha': 0.25, 'gamma': 2.0, 'loss_weight': 1.0, 'type': 'FocalLoss', 'use_sigmoid': True}, loss_shape={'beta': 0.2, 'loss_weight': 1.0, 'type': 'BoundedIoULoss'}, loss_cls={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_bbox={'beta': 1.0, 'loss_weight': 1.0, 'type': 'SmoothL1Loss'}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'conv_loc', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'})[source]

Guided-Anchor-based head (GA-RPN, GA-RetinaNet, etc.).

This GuidedAnchorHead will predict high-quality feature guided anchors and locations where anchors will be kept in inference. There are mainly 3 categories of bounding-boxes.

  • Sampled 9 pairs for target assignment. (approxes)
  • The square boxes where the predicted anchors are based on. (squares)
  • Guided anchors.

Please refer to https://arxiv.org/abs/1901.03278 for more details.

Parameters:
  • num_classes (int) – Number of classes.
  • in_channels (int) – Number of channels in the input feature map.
  • feat_channels (int) – Number of hidden channels.
  • approx_anchor_generator (dict) – Config dict for approx generator
  • square_anchor_generator (dict) – Config dict for square generator
  • anchor_coder (dict) – Config dict for anchor coder
  • bbox_coder (dict) – Config dict for bbox coder
  • reg_decoded_bbox (bool) – If true, the regression loss would be applied directly on decoded bounding boxes, converting both the predicted boxes and regression targets to absolute coordinates format. Default False. It should be True when using IoULoss, GIoULoss, or DIoULoss in the bbox head.
  • deform_groups – (int): Group number of DCN in FeatureAdaption module.
  • loc_filter_thr (float) – Threshold to filter out unconcerned regions.
  • loss_loc (dict) – Config of location loss.
  • loss_shape (dict) – Config of anchor shape loss.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox (dict) – Config of bbox regression loss.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:A tuple of classification scores and bbox prediction.
  • cls_scores (list[Tensor]): Classification scores for all scale levels, each is a 4D-tensor, the channels number is num_anchors * num_classes.
  • bbox_preds (list[Tensor]): Box energies / deltas for all scale levels, each is a 4D-tensor, the channels number is num_anchors * 4.
Return type:tuple
forward_single(x)[source]

Forward feature of a single scale level.

Parameters:x (Tensor) – Features of a single scale level.
Returns:cls_score (Tensor): Cls scores for a single scale level the channels number is num_anchors * num_classes. bbox_pred (Tensor): Box energies / deltas for a single scale level, the channels number is num_anchors * 4.
Return type:tuple
ga_loc_targets(gt_bboxes_list, featmap_sizes)[source]

Compute location targets for guided anchoring.

Each feature map is divided into positive, negative and ignore regions. - positive regions: target 1, weight 1 - ignore regions: target 0, weight 0 - negative regions: target 0, weight 0.1

Parameters:
  • gt_bboxes_list (list[Tensor]) – Gt bboxes of each image.
  • featmap_sizes (list[tuple]) – Multi level sizes of each feature maps.
Returns:

tuple

ga_shape_targets(approx_list, inside_flag_list, square_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, unmap_outputs=True)[source]

Compute guided anchoring targets.

Parameters:
  • approx_list (list[list]) – Multi level approxs of each image.
  • inside_flag_list (list[list]) – Multi level inside flags of each image.
  • square_list (list[list]) – Multi level squares of each image.
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • img_metas (list[dict]) – Meta info of each image.
  • gt_bboxes_ignore_list (list[Tensor]) – ignore list of gt bboxes.
  • unmap_outputs (bool) – unmap outputs or not.
Returns:

tuple

get_anchors(featmap_sizes, shape_preds, loc_preds, img_metas, use_loc_filter=False, device='cuda')[source]

Get squares according to feature map sizes and guided anchors.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • shape_preds (list[tensor]) – Multi-level shape predictions.
  • loc_preds (list[tensor]) – Multi-level location predictions.
  • img_metas (list[dict]) – Image meta info.
  • use_loc_filter (bool) – Use loc filter or not.
  • device (torch.device | str) – device for returned tensors
Returns:

square approxs of each image, guided anchors of each image,

loc masks of each image

Return type:

tuple

get_bboxes(cls_scores, bbox_preds, shape_preds, loc_preds, img_metas, cfg=None, rescale=False)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each level in the feature pyramid, has shape (N, num_anchors * num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box energies / deltas for each level in the feature pyramid, has shape (N, num_anchors * 4, H, W).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1. The shape of the second tensor in the tuple is (n,), and each element represents the class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

Example

>>> import mmcv
>>> self = AnchorHead(
>>>     num_classes=9,
>>>     in_channels=1,
>>>     anchor_generator=dict(
>>>         type='AnchorGenerator',
>>>         scales=[8],
>>>         ratios=[0.5, 1.0, 2.0],
>>>         strides=[4,]))
>>> img_metas = [{'img_shape': (32, 32, 3), 'scale_factor': 1}]
>>> cfg = mmcv.Config(dict(
>>>     score_thr=0.00,
>>>     nms=dict(type='nms', iou_thr=1.0),
>>>     max_per_img=10))
>>> feat = torch.rand(1, 1, 3, 3)
>>> cls_score, bbox_pred = self.forward_single(feat)
>>> # note the input lists are over different levels, not images
>>> cls_scores, bbox_preds = [cls_score], [bbox_pred]
>>> result_list = self.get_bboxes(cls_scores, bbox_preds,
>>>                               img_metas, cfg)
>>> det_bboxes, det_labels = result_list[0]
>>> assert len(result_list) == 1
>>> assert det_bboxes.shape[1] == 5
>>> assert len(det_bboxes) == len(det_labels) == cfg.max_per_img
get_sampled_approxs(featmap_sizes, img_metas, device='cuda')[source]

Get sampled approxs and inside flags according to feature map sizes.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • img_metas (list[dict]) – Image meta info.
  • device (torch.device | str) – device for returned tensors
Returns:

approxes of each image, inside flags of each image

Return type:

tuple

loss(cls_scores, bbox_preds, shape_preds, loc_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss. Default: None
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.FeatureAdaption(in_channels, out_channels, kernel_size=3, deform_groups=4, init_cfg={'layer': 'Conv2d', 'override': {'name': 'conv_adaption', 'std': 0.01, 'type': 'Normal'}, 'std': 0.1, 'type': 'Normal'})[source]

Feature Adaption Module.

Feature Adaption Module is implemented based on DCN v1. It uses anchor shape prediction rather than feature map to predict offsets of deform conv layer.

Parameters:
  • in_channels (int) – Number of channels in the input feature map.
  • out_channels (int) – Number of channels in the output feature map.
  • kernel_size (int) – Deformable conv kernel size.
  • deform_groups (int) – Deformable conv group size.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(x, shape)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mmdet.models.dense_heads.RPNHead(in_channels, init_cfg={'layer': 'Conv2d', 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

RPN head.

Parameters:
  • in_channels (int) – Number of channels in the input feature map.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward_single(x)[source]

Forward feature map of a single scale level.

get_bboxes(cls_scores, bbox_preds, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where the first 4 columns are bounding box positions (tl_x, tl_y, br_x, br_y) and the 5-th column is a score between 0 and 1. The second item is a (n,) tensor where each item is the predicted class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

loss(cls_scores, bbox_preds, gt_bboxes, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

onnx_export(x, img_metas)[source]

Test without augmentation.

Parameters:
  • x (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
  • img_metas (list[dict]) – Meta info of each image.
Returns:

dets of shape [N, num_det, 5]

and class labels of shape [N, num_det].

Return type:

tuple[Tensor, Tensor]

class mmdet.models.dense_heads.GARPNHead(in_channels, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'conv_loc', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

Guided-Anchor-based RPN head.

forward_single(x)[source]

Forward feature of a single scale level.

loss(cls_scores, bbox_preds, shape_preds, loc_preds, gt_bboxes, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss. Default: None
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.RetinaHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg=None, anchor_generator={'octave_base_scale': 4, 'ratios': [0.5, 1.0, 2.0], 'scales_per_octave': 3, 'strides': [8, 16, 32, 64, 128], 'type': 'AnchorGenerator'}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'retina_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

An anchor-based head used in RetinaNet.

The head contains two subnetworks. The first classifies anchor boxes and the second regresses deltas for the anchors.

Example

>>> import torch
>>> self = RetinaHead(11, 7)
>>> x = torch.rand(1, 7, 32, 32)
>>> cls_score, bbox_pred = self.forward_single(x)
>>> # Each anchor predicts a score for each class except background
>>> cls_per_anchor = cls_score.shape[1] / self.num_anchors
>>> box_per_anchor = bbox_pred.shape[1] / self.num_anchors
>>> assert cls_per_anchor == (self.num_classes)
>>> assert box_per_anchor == 4
forward_single(x)[source]

Forward feature of a single scale level.

Parameters:x (Tensor) – Features of a single scale level.
Returns:
cls_score (Tensor): Cls scores for a single scale level
the channels number is num_anchors * num_classes.
bbox_pred (Tensor): Box energies / deltas for a single scale
level, the channels number is num_anchors * 4.
Return type:tuple
class mmdet.models.dense_heads.RetinaSepBNHead(num_classes, num_ins, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg=None, init_cfg=None, **kwargs)[source]

“RetinaHead with separate BN.

In RetinaHead, conv/norm layers are shared across different FPN levels, while in RetinaSepBNHead, conv layers are shared across different FPN levels, but BN layers are separated.

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
Usually a tuple of classification scores and bbox prediction
cls_scores (list[Tensor]): Classification scores for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * 4.
Return type:tuple
init_weights()[source]

Initialize weights of the head.

class mmdet.models.dense_heads.GARetinaHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg=None, init_cfg=None, **kwargs)[source]

Guided-Anchor-based RetinaNet head.

forward_single(x)[source]

Forward feature map of a single scale level.

class mmdet.models.dense_heads.SSDHead(num_classes=80, in_channels=(512, 1024, 512, 256, 256, 256), stacked_convs=0, feat_channels=256, use_depthwise=False, conv_cfg=None, norm_cfg=None, act_cfg=None, anchor_generator={'basesize_ratio_range': (0.1, 0.9), 'input_size': 300, 'ratios': ([2], [2, 3], [2, 3], [2, 3], [2], [2]), 'scale_major': False, 'strides': [8, 16, 32, 64, 100, 300], 'type': 'SSDAnchorGenerator'}, bbox_coder={'clip_border': True, 'target_means': [0.0, 0.0, 0.0, 0.0], 'target_stds': [1.0, 1.0, 1.0, 1.0], 'type': 'DeltaXYWHBBoxCoder'}, reg_decoded_bbox=False, train_cfg=None, test_cfg=None, init_cfg={'bias': 0, 'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'})[source]

SSD head used in https://arxiv.org/abs/1512.02325.

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • stacked_convs (int) – Number of conv layers in cls and reg tower. Default: 0.
  • feat_channels (int) – Number of hidden channels when stacked_convs > 0. Default: 256.
  • use_depthwise (bool) – Whether to use DepthwiseSeparableConv. Default: False.
  • conv_cfg (dict) – Dictionary to construct and config conv layer. Default: None.
  • norm_cfg (dict) – Dictionary to construct and config norm layer. Default: None.
  • act_cfg (dict) – Dictionary to construct and config activation layer. Default: None.
  • anchor_generator (dict) – Config dict for anchor generator
  • bbox_coder (dict) – Config of bounding box coder.
  • reg_decoded_bbox (bool) – If true, the regression loss would be applied directly on decoded bounding boxes, converting both the predicted boxes and regression targets to absolute coordinates format. Default False. It should be True when using IoULoss, GIoULoss, or DIoULoss in the bbox head.
  • train_cfg (dict) – Training config of anchor head.
  • test_cfg (dict) – Testing config of anchor head.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
cls_scores (list[Tensor]): Classification scores for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * 4.
Return type:tuple
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

loss_single(cls_score, bbox_pred, anchor, labels, label_weights, bbox_targets, bbox_weights, num_total_samples)[source]

Compute loss of a single image.

Parameters:
  • cls_score (Tensor) – Box scores for eachimage Has shape (num_total_anchors, num_classes).
  • bbox_pred (Tensor) – Box energies / deltas for each image level with shape (num_total_anchors, 4).
  • anchors (Tensor) – Box reference for each scale level with shape (num_total_anchors, 4).
  • labels (Tensor) – Labels of each anchors with shape (num_total_anchors,).
  • label_weights (Tensor) – Label weights of each anchor with shape (num_total_anchors,)
  • bbox_targets (Tensor) – BBox regression targets of each anchor wight shape (num_total_anchors, 4).
  • bbox_weights (Tensor) – BBox regression loss weights of each anchor with shape (num_total_anchors, 4).
  • num_total_samples (int) – If sampling, num total samples equal to the number of total anchors; Otherwise, it is the number of positive anchors.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.FCOSHead(num_classes, in_channels, regress_ranges=((-1, 64), (64, 128), (128, 256), (256, 512), (512, 100000000.0)), center_sampling=False, center_sample_radius=1.5, norm_on_bbox=False, centerness_on_reg=False, loss_cls={'alpha': 0.25, 'gamma': 2.0, 'loss_weight': 1.0, 'type': 'FocalLoss', 'use_sigmoid': True}, loss_bbox={'loss_weight': 1.0, 'type': 'IoULoss'}, loss_centerness={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, norm_cfg={'num_groups': 32, 'requires_grad': True, 'type': 'GN'}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'conv_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

Anchor-free head used in FCOS.

The FCOS head does not use anchor boxes. Instead bounding boxes are predicted at each pixel and a centerness measure is used to suppress low-quality predictions. Here norm_on_bbox, centerness_on_reg, dcn_on_last_conv are training tricks used in official repo, which will bring remarkable mAP gains of up to 4.9. Please see https://github.com/tianzhi0549/FCOS for more detail.

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • strides (list[int] | list[tuple[int, int]]) – Strides of points in multiple feature levels. Default: (4, 8, 16, 32, 64).
  • regress_ranges (tuple[tuple[int, int]]) – Regress range of multiple level points.
  • center_sampling (bool) – If true, use center sampling. Default: False.
  • center_sample_radius (float) – Radius of center sampling. Default: 1.5.
  • norm_on_bbox (bool) – If true, normalize the regression targets with FPN strides. Default: False.
  • centerness_on_reg (bool) – If true, position centerness on the regress branch. Please refer to https://github.com/tianzhi0549/FCOS/issues/89#issuecomment-516877042. Default: False.
  • conv_bias (bool | str) – If specified as auto, it will be decided by the norm_cfg. Bias of conv will be set as True if norm_cfg is None, otherwise False. Default: “auto”.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox (dict) – Config of localization loss.
  • loss_centerness (dict) – Config of centerness loss.
  • norm_cfg (dict) – dictionary to construct and config norm layer. Default: norm_cfg=dict(type=’GN’, num_groups=32, requires_grad=True).
  • init_cfg (dict or list[dict], optional) – Initialization config dict.

Example

>>> self = FCOSHead(11, 7)
>>> feats = [torch.rand(1, 7, s, s) for s in [4, 8, 16, 32, 64]]
>>> cls_score, bbox_pred, centerness = self.forward(feats)
>>> assert len(cls_score) == len(self.scales)
centerness_target(pos_bbox_targets)[source]

Compute centerness targets.

Parameters:pos_bbox_targets (Tensor) – BBox targets of positive bboxes in shape (num_pos, 4)
Returns:Centerness target.
Return type:Tensor
forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:cls_scores (list[Tensor]): Box scores for each scale level, each is a 4D-tensor, the channel number is num_points * num_classes. bbox_preds (list[Tensor]): Box energies / deltas for each scale level, each is a 4D-tensor, the channel number is num_points * 4. centernesses (list[Tensor]): centerness for each scale level, each is a 4D-tensor, the channel number is num_points * 1.
Return type:tuple
forward_single(x, scale, stride)[source]

Forward features of a single scale level.

Parameters:
  • x (Tensor) – FPN feature maps of the specified stride.
  • ( (scale) – obj: mmcv.cnn.Scale): Learnable scale module to resize the bbox prediction.
  • stride (int) – The corresponding stride for feature maps, only used to normalize the bbox prediction when self.norm_on_bbox is True.
Returns:

scores for each class, bbox predictions and centerness predictions of input feature maps.

Return type:

tuple

get_bboxes(cls_scores, bbox_preds, centernesses, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level with shape (N, num_points * num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_points * 4, H, W).
  • centernesses (list[Tensor]) – Centerness for each scale level with shape (N, num_points * 1, H, W).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used. Default: None.
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1. The shape of the second tensor in the tuple is (n,), and each element represents the class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

get_targets(points, gt_bboxes_list, gt_labels_list)[source]

Compute regression, classification and centerness targets for points in multiple images.

Parameters:
  • points (list[Tensor]) – Points of each fpn level, each has shape (num_points, 2).
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image, each has shape (num_gt, 4).
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box, each has shape (num_gt,).
Returns:

concat_lvl_labels (list[Tensor]): Labels of each level. concat_lvl_bbox_targets (list[Tensor]): BBox targets of each level.

Return type:

tuple

loss(cls_scores, bbox_preds, centernesses, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level, each is a 4D-tensor, the channel number is num_points * num_classes.
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level, each is a 4D-tensor, the channel number is num_points * 4.
  • centernesses (list[Tensor]) – centerness for each scale level, each is a 4D-tensor, the channel number is num_points * 1.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.RepPointsHead(num_classes, in_channels, point_feat_channels=256, num_points=9, gradient_mul=0.1, point_strides=[8, 16, 32, 64, 128], point_base_scale=4, loss_cls={'alpha': 0.25, 'gamma': 2.0, 'loss_weight': 1.0, 'type': 'FocalLoss', 'use_sigmoid': True}, loss_bbox_init={'beta': 0.1111111111111111, 'loss_weight': 0.5, 'type': 'SmoothL1Loss'}, loss_bbox_refine={'beta': 0.1111111111111111, 'loss_weight': 1.0, 'type': 'SmoothL1Loss'}, use_grid_points=False, center_init=True, transform_method='moment', moment_mul=0.01, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'reppoints_cls_out', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

RepPoint head.

Parameters:
  • point_feat_channels (int) – Number of channels of points features.
  • gradient_mul (float) – The multiplier to gradients from points refinement and recognition.
  • point_strides (Iterable) – points strides.
  • point_base_scale (int) – bbox scale for assigning labels.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox_init (dict) – Config of initial points loss.
  • loss_bbox_refine (dict) – Config of points loss in refinement.
  • use_grid_points (bool) – If we use bounding box representation, the
  • is represented as grid points on the bounding box. (reppoints) –
  • center_init (bool) – Whether to use center point assignment.
  • transform_method (str) – The methods to transform RepPoints to bbox.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
centers_to_bboxes(point_list)[source]

Get bboxes according to center points.

Only used in MaxIoUAssigner.

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
Usually contain classification scores and bbox predictions.
cls_scores (list[Tensor]): Box scores for each scale level,
each is a 4D-tensor, the channel number is num_points * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for each scale
level, each is a 4D-tensor, the channel number is num_points * 4.
Return type:tuple
forward_single(x)[source]

Forward feature map of a single FPN level.

gen_grid_from_reg(reg, previous_boxes)[source]

Base on the previous bboxes and regression values, we compute the regressed bboxes and generate the grids on the bboxes.

Parameters:
  • reg – the regression value to previous bboxes.
  • previous_boxes – previous bboxes.
Returns:

generate grids on the regressed bboxes.

get_bboxes(cls_scores, pts_preds_init, pts_preds_refine, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_points * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_points * 4, H, W)
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space
get_points(featmap_sizes, img_metas, device)[source]

Get points according to feature map sizes.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • img_metas (list[dict]) – Image meta info.
Returns:

points of each image, valid flags of each image

Return type:

tuple

get_targets(proposals_list, valid_flag_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, stage='init', label_channels=1, unmap_outputs=True)[source]

Compute corresponding GT box and classification targets for proposals.

Parameters:
  • proposals_list (list[list]) – Multi level points/bboxes of each image.
  • valid_flag_list (list[list]) – Multi level valid flags of each image.
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • img_metas (list[dict]) – Meta info of each image.
  • gt_bboxes_ignore_list (list[Tensor]) – Ground truth bboxes to be ignored.
  • gt_bboxes_list – Ground truth labels of each box.
  • stage (str) – init or refine. Generate target for init stage or refine stage
  • label_channels (int) – Channel of label.
  • unmap_outputs (bool) – Whether to map outputs back to the original set of anchors.
Returns:

  • labels_list (list[Tensor]): Labels of each level.
  • label_weights_list (list[Tensor]): Label weights of each level. # noqa: E501
  • bbox_gt_list (list[Tensor]): Ground truth bbox of each level.
  • proposal_list (list[Tensor]): Proposals(points/bboxes) of each level. # noqa: E501
  • proposal_weights_list (list[Tensor]): Proposal weights of each level. # noqa: E501
  • num_total_pos (int): Number of positive samples in all images. # noqa: E501
  • num_total_neg (int): Number of negative samples in all images. # noqa: E501

Return type:

tuple

loss(cls_scores, pts_preds_init, pts_preds_refine, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level, each is a 4D-tensor, the channel number is num_points * num_classes.
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level, each is a 4D-tensor, the channel number is num_points * 4.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
offset_to_pts(center_list, pred_list)[source]

Change from point offset to point coordinate.

points2bbox(pts, y_first=True)[source]

Converting the points set into bounding box.

Parameters:
  • pts – the input points sets (fields), each points set (fields) is represented as 2n scalar.
  • y_first – if y_first=True, the point set is represented as [y1, x1, y2, x2 … yn, xn], otherwise the point set is represented as [x1, y1, x2, y2 … xn, yn].
Returns:

each points set is converting to a bbox [x1, y1, x2, y2].

class mmdet.models.dense_heads.FoveaHead(num_classes, in_channels, base_edge_list=(16, 32, 64, 128, 256), scale_ranges=((8, 32), (16, 64), (32, 128), (64, 256), (128, 512)), sigma=0.4, with_deform=False, deform_groups=4, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'conv_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

FoveaBox: Beyond Anchor-based Object Detector https://arxiv.org/abs/1904.03797

forward_single(x)[source]

Forward features of a single scale level.

Parameters:x (Tensor) – FPN feature maps of the specified stride.
Returns:
Scores for each class, bbox predictions, features
after classification and regression conv layers, some models needs these features like FCOS.
Return type:tuple
get_bboxes(cls_scores, bbox_preds, img_metas, cfg=None, rescale=None)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_points * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_points * 4, H, W)
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space
get_targets(gt_bbox_list, gt_label_list, featmap_sizes, points)[source]

Compute regression, classification and centerness targets for points in multiple images.

Parameters:
  • points (list[Tensor]) – Points of each fpn level, each has shape (num_points, 2).
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image, each has shape (num_gt, 4).
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box, each has shape (num_gt,).
loss(cls_scores, bbox_preds, gt_bbox_list, gt_label_list, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level, each is a 4D-tensor, the channel number is num_points * num_classes.
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level, each is a 4D-tensor, the channel number is num_points * 4.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
class mmdet.models.dense_heads.FreeAnchorRetinaHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg=None, pre_anchor_topk=50, bbox_thr=0.6, gamma=2.0, alpha=0.5, **kwargs)[source]

FreeAnchor RetinaHead used in https://arxiv.org/abs/1909.02466.

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • stacked_convs (int) – Number of conv layers in cls and reg tower. Default: 4.
  • conv_cfg (dict) – dictionary to construct and config conv layer. Default: None.
  • norm_cfg (dict) – dictionary to construct and config norm layer. Default: norm_cfg=dict(type=’GN’, num_groups=32, requires_grad=True).
  • pre_anchor_topk (int) – Number of boxes that be token in each bag.
  • bbox_thr (float) – The threshold of the saturated linear function. It is usually the same with the IoU threshold used in NMS.
  • gamma (float) – Gamma parameter in focal loss.
  • alpha (float) – Alpha parameter in focal loss.
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

negative_bag_loss(cls_prob, box_prob)[source]

Compute negative bag loss.

\(FL((1 - P_{a_{j} \in A_{+}}) * (1 - P_{j}^{bg}))\).

\(P_{a_{j} \in A_{+}}\): Box_probability of matched samples.

\(P_{j}^{bg}\): Classification probability of negative samples.

Parameters:
  • cls_prob (Tensor) – Classification probability, in shape (num_img, num_anchors, num_classes).
  • box_prob (Tensor) – Box probability, in shape (num_img, num_anchors, num_classes).
Returns:

Negative bag loss in shape (num_img, num_anchors, num_classes).

Return type:

Tensor

positive_bag_loss(matched_cls_prob, matched_box_prob)[source]

Compute positive bag loss.

\(-log( Mean-max(P_{ij}^{cls} * P_{ij}^{loc}) )\).

\(P_{ij}^{cls}\): matched_cls_prob, classification probability of matched samples.

\(P_{ij}^{loc}\): matched_box_prob, box probability of matched samples.

Parameters:
  • matched_cls_prob (Tensor) – Classification probability of matched samples in shape (num_gt, pre_anchor_topk).
  • matched_box_prob (Tensor) – BBox probability of matched samples, in shape (num_gt, pre_anchor_topk).
Returns:

Positive bag loss in shape (num_gt,).

Return type:

Tensor

class mmdet.models.dense_heads.ATSSHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg={'num_groups': 32, 'requires_grad': True, 'type': 'GN'}, loss_centerness={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'atss_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection.

ATSS head structure is similar with FCOS, however ATSS use anchor boxes and assign label by Adaptive Training Sample Selection instead max-iou.

https://arxiv.org/abs/1912.02424

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
Usually a tuple of classification scores and bbox prediction
cls_scores (list[Tensor]): Classification scores for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for all scale
levels, each is a 4D-tensor, the channels number is num_anchors * 4.
Return type:tuple
forward_single(x, scale)[source]

Forward feature of a single scale level.

Parameters:
  • x (Tensor) – Features of a single scale level.
  • ( (scale) – obj: mmcv.cnn.Scale): Learnable scale module to resize the bbox prediction.
Returns:

cls_score (Tensor): Cls scores for a single scale level

the channels number is num_anchors * num_classes.

bbox_pred (Tensor): Box energies / deltas for a single scale

level, the channels number is num_anchors * 4.

centerness (Tensor): Centerness for a single scale level, the

channel number is (N, num_anchors * 1, H, W).

Return type:

tuple

get_bboxes(cls_scores, bbox_preds, centernesses, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level with shape (N, num_anchors * num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W).
  • centernesses (list[Tensor]) – Centerness for each scale level with shape (N, num_anchors * 1, H, W).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used. Default: None.
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1. The shape of the second tensor in the tuple is (n,), and each element represents the class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

get_targets(anchor_list, valid_flag_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, label_channels=1, unmap_outputs=True)[source]

Get targets for ATSS head.

This method is almost the same as AnchorHead.get_targets(). Besides returning the targets as the parent method does, it also returns the anchors as the first element of the returned tuple.

loss(cls_scores, bbox_preds, centernesses, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • centernesses (list[Tensor]) – Centerness for each scale level with shape (N, num_anchors * 1, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor] | None) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

loss_single(anchors, cls_score, bbox_pred, centerness, labels, label_weights, bbox_targets, num_total_samples)[source]

Compute loss of a single scale level.

Parameters:
  • cls_score (Tensor) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W).
  • bbox_pred (Tensor) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W).
  • anchors (Tensor) – Box reference for each scale level with shape (N, num_total_anchors, 4).
  • labels (Tensor) – Labels of each anchors with shape (N, num_total_anchors).
  • label_weights (Tensor) – Label weights of each anchor with shape (N, num_total_anchors)
  • bbox_targets (Tensor) – BBox regression targets of each anchor wight shape (N, num_total_anchors, 4).
  • num_total_samples (int) – Number os positive samples that is reduced over all GPUs.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.FSAFHead(*args, score_threshold=None, init_cfg=None, **kwargs)[source]

Anchor-free head used in FSAF.

The head contains two subnetworks. The first classifies anchor boxes and the second regresses deltas for the anchors (num_anchors is 1 for anchor- free methods)

Parameters:
  • *args – Same as its base class in RetinaHead
  • score_threshold (float, optional) – The score_threshold to calculate positive recall. If given, prediction scores lower than this value is counted as incorrect prediction. Default to None.
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
  • **kwargs – Same as its base class in RetinaHead

Example

>>> import torch
>>> self = FSAFHead(11, 7)
>>> x = torch.rand(1, 7, 32, 32)
>>> cls_score, bbox_pred = self.forward_single(x)
>>> # Each anchor predicts a score for each class except background
>>> cls_per_anchor = cls_score.shape[1] / self.num_anchors
>>> box_per_anchor = bbox_pred.shape[1] / self.num_anchors
>>> assert cls_per_anchor == self.num_classes
>>> assert box_per_anchor == 4
calculate_pos_recall(cls_scores, labels_list, pos_inds)[source]

Calculate positive recall with score threshold.

Parameters:
  • cls_scores (list[Tensor]) – Classification scores at all fpn levels. Each tensor is in shape (N, num_classes * num_anchors, H, W)
  • labels_list (list[Tensor]) – The label that each anchor is assigned to. Shape (N * H * W * num_anchors, )
  • pos_inds (list[Tensor]) – List of bool tensors indicating whether the anchor is assigned to a positive label. Shape (N * H * W * num_anchors, )
Returns:

A single float number indicating the positive recall.

Return type:

Tensor

collect_loss_level_single(cls_loss, reg_loss, assigned_gt_inds, labels_seq)[source]

Get the average loss in each FPN level w.r.t. each gt label.

Parameters:
  • cls_loss (Tensor) – Classification loss of each feature map pixel, shape (num_anchor, num_class)
  • reg_loss (Tensor) – Regression loss of each feature map pixel, shape (num_anchor, 4)
  • assigned_gt_inds (Tensor) – It indicates which gt the prior is assigned to (0-based, -1: no assignment). shape (num_anchor),
  • labels_seq – The rank of labels. shape (num_gt)
Returns:

(num_gt), average loss of each gt in this level

Return type:

shape

forward_single(x)[source]

Forward feature map of a single scale level.

Parameters:x (Tensor) – Feature map of a single scale level.
Returns:
cls_score (Tensor): Box scores for each scale level
Has shape (N, num_points * num_classes, H, W).
bbox_pred (Tensor): Box energies / deltas for each scale
level with shape (N, num_points * 4, H, W).
Return type:tuple (Tensor)
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_points * num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_points * 4, H, W).
  • gt_bboxes (list[Tensor]) – each item are the truth boxes for each image in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

reweight_loss_single(cls_loss, reg_loss, assigned_gt_inds, labels, level, min_levels)[source]

Reweight loss values at each level.

Reassign loss values at each level by masking those where the pre-calculated loss is too large. Then return the reduced losses.

Parameters:
  • cls_loss (Tensor) – Element-wise classification loss. Shape: (num_anchors, num_classes)
  • reg_loss (Tensor) – Element-wise regression loss. Shape: (num_anchors, 4)
  • assigned_gt_inds (Tensor) – The gt indices that each anchor bbox is assigned to. -1 denotes a negative anchor, otherwise it is the gt index (0-based). Shape: (num_anchors, ),
  • labels (Tensor) – Label assigned to anchors. Shape: (num_anchors, ).
  • level (int) – The current level index in the pyramid (0-4 for RetinaNet)
  • min_levels (Tensor) – The best-matching level for each gt. Shape: (num_gts, ),
Returns:

  • cls_loss: Reduced corrected classification loss. Scalar.
  • reg_loss: Reduced corrected regression loss. Scalar.
  • pos_flags (Tensor): Corrected bool tensor indicating the final positive anchors. Shape: (num_anchors, ).

Return type:

tuple

class mmdet.models.dense_heads.NASFCOSHead(*args, init_cfg=None, **kwargs)[source]

Anchor-free head used in NASFCOS.

It is quite similar with FCOS head, except for the searched structure of classification branch and bbox regression branch, where a structure of “dconv3x3, conv3x3, dconv3x3, conv1x1” is utilized instead.

class mmdet.models.dense_heads.PISARetinaHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg=None, anchor_generator={'octave_base_scale': 4, 'ratios': [0.5, 1.0, 2.0], 'scales_per_octave': 3, 'strides': [8, 16, 32, 64, 128], 'type': 'AnchorGenerator'}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'retina_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

PISA Retinanet Head.

The head owns the same structure with Retinanet Head, but differs in two

aspects: 1. Importance-based Sample Reweighting Positive (ISR-P) is applied to

change the positive loss weights.
  1. Classification-aware regression loss is adopted as a third loss.
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes of each image with shape (num_obj, 4).
  • gt_labels (list[Tensor]) – Ground truth labels of each image with shape (num_obj, 4).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor]) – Ignored gt bboxes of each image. Default: None.
Returns:

Loss dict, comprise classification loss, regression loss and

carl loss.

Return type:

dict

class mmdet.models.dense_heads.PISASSDHead(num_classes=80, in_channels=(512, 1024, 512, 256, 256, 256), stacked_convs=0, feat_channels=256, use_depthwise=False, conv_cfg=None, norm_cfg=None, act_cfg=None, anchor_generator={'basesize_ratio_range': (0.1, 0.9), 'input_size': 300, 'ratios': ([2], [2, 3], [2, 3], [2, 3], [2], [2]), 'scale_major': False, 'strides': [8, 16, 32, 64, 100, 300], 'type': 'SSDAnchorGenerator'}, bbox_coder={'clip_border': True, 'target_means': [0.0, 0.0, 0.0, 0.0], 'target_stds': [1.0, 1.0, 1.0, 1.0], 'type': 'DeltaXYWHBBoxCoder'}, reg_decoded_bbox=False, train_cfg=None, test_cfg=None, init_cfg={'bias': 0, 'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'})[source]
loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes of each image with shape (num_obj, 4).
  • gt_labels (list[Tensor]) – Ground truth labels of each image with shape (num_obj, 4).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor]) – Ignored gt bboxes of each image. Default: None.
Returns:

Loss dict, comprise classification loss regression loss and

carl loss.

Return type:

dict

class mmdet.models.dense_heads.GFLHead(num_classes, in_channels, stacked_convs=4, conv_cfg=None, norm_cfg={'num_groups': 32, 'requires_grad': True, 'type': 'GN'}, loss_dfl={'loss_weight': 0.25, 'type': 'DistributionFocalLoss'}, reg_max=16, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'gfl_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'}, **kwargs)[source]

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection.

GFL head structure is similar with ATSS, however GFL uses 1) joint representation for classification and localization quality, and 2) flexible General distribution for bounding box locations, which are supervised by Quality Focal Loss (QFL) and Distribution Focal Loss (DFL), respectively

https://arxiv.org/abs/2006.04388

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • stacked_convs (int) – Number of conv layers in cls and reg tower. Default: 4.
  • conv_cfg (dict) – dictionary to construct and config conv layer. Default: None.
  • norm_cfg (dict) – dictionary to construct and config norm layer. Default: dict(type=’GN’, num_groups=32, requires_grad=True).
  • loss_qfl (dict) – Config of Quality Focal Loss (QFL).
  • reg_max (int) – Max value of integral set :math: {0, …, reg_max} in QFL setting. Default: 16.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.

Example

>>> self = GFLHead(11, 7)
>>> feats = [torch.rand(1, 7, s, s) for s in [4, 8, 16, 32, 64]]
>>> cls_quality_score, bbox_pred = self.forward(feats)
>>> assert len(cls_quality_score) == len(self.scales)
anchor_center(anchors)[source]

Get anchor centers from anchors.

Parameters:anchors (Tensor) – Anchor list with shape (N, 4), “xyxy” format.
Returns:Anchor centers with shape (N, 2), “xy” format.
Return type:Tensor
forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
Usually a tuple of classification scores and bbox prediction
cls_scores (list[Tensor]): Classification and quality (IoU)
joint scores for all scale levels, each is a 4D-tensor, the channel number is num_classes.
bbox_preds (list[Tensor]): Box distribution logits for all
scale levels, each is a 4D-tensor, the channel number is 4*(n+1), n is max value of integral set.
Return type:tuple
forward_single(x, scale)[source]

Forward feature of a single scale level.

Parameters:
  • x (Tensor) – Features of a single scale level.
  • ( (scale) – obj: mmcv.cnn.Scale): Learnable scale module to resize the bbox prediction.
Returns:

cls_score (Tensor): Cls and quality joint scores for a single

scale level the channel number is num_classes.

bbox_pred (Tensor): Box distribution logits for a single scale

level, the channel number is 4*(n+1), n is max value of integral set.

Return type:

tuple

get_targets(anchor_list, valid_flag_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, label_channels=1, unmap_outputs=True)[source]

Get targets for GFL head.

This method is almost the same as AnchorHead.get_targets(). Besides returning the targets as the parent method does, it also returns the anchors as the first element of the returned tuple.

loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Cls and quality scores for each scale level has shape (N, num_classes, H, W).
  • bbox_preds (list[Tensor]) – Box distribution logits for each scale level with shape (N, 4*(n+1), H, W), n is max value of integral set.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor] | None) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

loss_single(anchors, cls_score, bbox_pred, labels, label_weights, bbox_targets, stride, num_total_samples)[source]

Compute loss of a single scale level.

Parameters:
  • anchors (Tensor) – Box reference for each scale level with shape (N, num_total_anchors, 4).
  • cls_score (Tensor) – Cls and quality joint scores for each scale level has shape (N, num_classes, H, W).
  • bbox_pred (Tensor) – Box distribution logits for each scale level with shape (N, 4*(n+1), H, W), n is max value of integral set.
  • labels (Tensor) – Labels of each anchors with shape (N, num_total_anchors).
  • label_weights (Tensor) – Label weights of each anchor with shape (N, num_total_anchors)
  • bbox_targets (Tensor) – BBox regression targets of each anchor wight shape (N, num_total_anchors, 4).
  • stride (tuple) – Stride in this scale level.
  • num_total_samples (int) – Number of positive samples that is reduced over all GPUs.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

class mmdet.models.dense_heads.CornerHead(num_classes, in_channels, num_feat_levels=2, corner_emb_channels=1, train_cfg=None, test_cfg=None, loss_heatmap={'alpha': 2.0, 'gamma': 4.0, 'loss_weight': 1, 'type': 'GaussianFocalLoss'}, loss_embedding={'pull_weight': 0.25, 'push_weight': 0.25, 'type': 'AssociativeEmbeddingLoss'}, loss_offset={'beta': 1.0, 'loss_weight': 1, 'type': 'SmoothL1Loss'}, init_cfg=None)[source]

Head of CornerNet: Detecting Objects as Paired Keypoints.

Code is modified from the official github repo .

More details can be found in the paper .

Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • num_feat_levels (int) – Levels of feature from the previous module. 2 for HourglassNet-104 and 1 for HourglassNet-52. Because HourglassNet-104 outputs the final feature and intermediate supervision feature and HourglassNet-52 only outputs the final feature. Default: 2.
  • corner_emb_channels (int) – Channel of embedding vector. Default: 1.
  • train_cfg (dict | None) – Training config. Useless in CornerHead, but we keep this variable for SingleStageDetector. Default: None.
  • test_cfg (dict | None) – Testing config of CornerHead. Default: None.
  • loss_heatmap (dict | None) – Config of corner heatmap loss. Default: GaussianFocalLoss.
  • loss_embedding (dict | None) – Config of corner embedding loss. Default: AssociativeEmbeddingLoss.
  • loss_offset (dict | None) – Config of corner offset loss. Default: SmoothL1Loss.
  • init_cfg (dict or list[dict], optional) – Initialization config dict. Default: None
decode_heatmap(tl_heat, br_heat, tl_off, br_off, tl_emb=None, br_emb=None, tl_centripetal_shift=None, br_centripetal_shift=None, img_meta=None, k=100, kernel=3, distance_threshold=0.5, num_dets=1000)[source]

Transform outputs for a single batch item into raw bbox predictions.

Parameters:
  • tl_heat (Tensor) – Top-left corner heatmap for current level with shape (N, num_classes, H, W).
  • br_heat (Tensor) – Bottom-right corner heatmap for current level with shape (N, num_classes, H, W).
  • tl_off (Tensor) – Top-left corner offset for current level with shape (N, corner_offset_channels, H, W).
  • br_off (Tensor) – Bottom-right corner offset for current level with shape (N, corner_offset_channels, H, W).
  • tl_emb (Tensor | None) – Top-left corner embedding for current level with shape (N, corner_emb_channels, H, W).
  • br_emb (Tensor | None) – Bottom-right corner embedding for current level with shape (N, corner_emb_channels, H, W).
  • tl_centripetal_shift (Tensor | None) – Top-left centripetal shift for current level with shape (N, 2, H, W).
  • br_centripetal_shift (Tensor | None) – Bottom-right centripetal shift for current level with shape (N, 2, H, W).
  • img_meta (dict) – Meta information of current image, e.g., image size, scaling factor, etc.
  • k (int) – Get top k corner keypoints from heatmap.
  • kernel (int) – Max pooling kernel for extract local maximum pixels.
  • distance_threshold (float) – Distance threshold. Top-left and bottom-right corner keypoints with feature distance less than the threshold will be regarded as keypoints from same object.
  • num_dets (int) – Num of raw boxes before doing nms.
Returns:

Decoded output of CornerHead, containing the following Tensors:

  • bboxes (Tensor): Coords of each box.
  • scores (Tensor): Scores of each box.
  • clses (Tensor): Categories of each box.

Return type:

tuple[torch.Tensor]

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:Usually a tuple of corner heatmaps, offset heatmaps and embedding heatmaps.
  • tl_heats (list[Tensor]): Top-left corner heatmaps for all levels, each is a 4D-tensor, the channels number is num_classes.
  • br_heats (list[Tensor]): Bottom-right corner heatmaps for all levels, each is a 4D-tensor, the channels number is num_classes.
  • tl_embs (list[Tensor] | list[None]): Top-left embedding heatmaps for all levels, each is a 4D-tensor or None. If not None, the channels number is corner_emb_channels.
  • br_embs (list[Tensor] | list[None]): Bottom-right embedding heatmaps for all levels, each is a 4D-tensor or None. If not None, the channels number is corner_emb_channels.
  • tl_offs (list[Tensor]): Top-left offset heatmaps for all levels, each is a 4D-tensor. The channels number is corner_offset_channels.
  • br_offs (list[Tensor]): Bottom-right offset heatmaps for all levels, each is a 4D-tensor. The channels number is corner_offset_channels.
Return type:tuple
forward_single(x, lvl_ind, return_pool=False)[source]

Forward feature of a single level.

Parameters:
  • x (Tensor) – Feature of a single level.
  • lvl_ind (int) – Level index of current feature.
  • return_pool (bool) – Return corner pool feature or not.
Returns:

A tuple of CornerHead’s output for current feature level. Containing the following Tensors:

  • tl_heat (Tensor): Predicted top-left corner heatmap.
  • br_heat (Tensor): Predicted bottom-right corner heatmap.
  • tl_emb (Tensor | None): Predicted top-left embedding heatmap. None for self.with_corner_emb == False.
  • br_emb (Tensor | None): Predicted bottom-right embedding heatmap. None for self.with_corner_emb == False.
  • tl_off (Tensor): Predicted top-left offset heatmap.
  • br_off (Tensor): Predicted bottom-right offset heatmap.
  • tl_pool (Tensor): Top-left corner pool feature. Not must have.
  • br_pool (Tensor): Bottom-right corner pool feature. Not must have.

Return type:

tuple[Tensor]

get_bboxes(tl_heats, br_heats, tl_embs, br_embs, tl_offs, br_offs, img_metas, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • tl_heats (list[Tensor]) – Top-left corner heatmaps for each level with shape (N, num_classes, H, W).
  • br_heats (list[Tensor]) – Bottom-right corner heatmaps for each level with shape (N, num_classes, H, W).
  • tl_embs (list[Tensor]) – Top-left corner embeddings for each level with shape (N, corner_emb_channels, H, W).
  • br_embs (list[Tensor]) – Bottom-right corner embeddings for each level with shape (N, corner_emb_channels, H, W).
  • tl_offs (list[Tensor]) – Top-left corner offsets for each level with shape (N, corner_offset_channels, H, W).
  • br_offs (list[Tensor]) – Bottom-right corner offsets for each level with shape (N, corner_offset_channels, H, W).
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
get_targets(gt_bboxes, gt_labels, feat_shape, img_shape, with_corner_emb=False, with_guiding_shift=False, with_centripetal_shift=False)[source]

Generate corner targets.

Including corner heatmap, corner offset.

Optional: corner embedding, corner guiding shift, centripetal shift.

For CornerNet, we generate corner heatmap, corner offset and corner embedding from this function.

For CentripetalNet, we generate corner heatmap, corner offset, guiding shift and centripetal shift from this function.

Parameters:
  • gt_bboxes (list[Tensor]) – Ground truth bboxes of each image, each has shape (num_gt, 4).
  • gt_labels (list[Tensor]) – Ground truth labels of each box, each has shape (num_gt,).
  • feat_shape (list[int]) – Shape of output feature, [batch, channel, height, width].
  • img_shape (list[int]) – Shape of input image, [height, width, channel].
  • with_corner_emb (bool) – Generate corner embedding target or not. Default: False.
  • with_guiding_shift (bool) – Generate guiding shift target or not. Default: False.
  • with_centripetal_shift (bool) – Generate centripetal shift target or not. Default: False.
Returns:

Ground truth of corner heatmap, corner offset, corner embedding, guiding shift and centripetal shift. Containing the following keys:

  • topleft_heatmap (Tensor): Ground truth top-left corner heatmap.
  • bottomright_heatmap (Tensor): Ground truth bottom-right corner heatmap.
  • topleft_offset (Tensor): Ground truth top-left corner offset.
  • bottomright_offset (Tensor): Ground truth bottom-right corner offset.
  • corner_embedding (list[list[list[int]]]): Ground truth corner embedding. Not must have.
  • topleft_guiding_shift (Tensor): Ground truth top-left corner guiding shift. Not must have.
  • bottomright_guiding_shift (Tensor): Ground truth bottom-right corner guiding shift. Not must have.
  • topleft_centripetal_shift (Tensor): Ground truth top-left corner centripetal shift. Not must have.
  • bottomright_centripetal_shift (Tensor): Ground truth bottom-right corner centripetal shift. Not must have.

Return type:

dict

init_weights()[source]

Initialize the weights.

loss(tl_heats, br_heats, tl_embs, br_embs, tl_offs, br_offs, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • tl_heats (list[Tensor]) – Top-left corner heatmaps for each level with shape (N, num_classes, H, W).
  • br_heats (list[Tensor]) – Bottom-right corner heatmaps for each level with shape (N, num_classes, H, W).
  • tl_embs (list[Tensor]) – Top-left corner embeddings for each level with shape (N, corner_emb_channels, H, W).
  • br_embs (list[Tensor]) – Bottom-right corner embeddings for each level with shape (N, corner_emb_channels, H, W).
  • tl_offs (list[Tensor]) – Top-left corner offsets for each level with shape (N, corner_offset_channels, H, W).
  • br_offs (list[Tensor]) – Bottom-right corner offsets for each level with shape (N, corner_offset_channels, H, W).
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [left, top, right, bottom] format.
  • gt_labels (list[Tensor]) – Class indices corresponding to each box.
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor] | None) – Specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components. Containing the following losses:

  • det_loss (list[Tensor]): Corner keypoint losses of all feature levels.
  • pull_loss (list[Tensor]): Part one of AssociativeEmbedding losses of all feature levels.
  • push_loss (list[Tensor]): Part two of AssociativeEmbedding losses of all feature levels.
  • off_loss (list[Tensor]): Corner offset losses of all feature levels.

Return type:

dict[str, Tensor]

loss_single(tl_hmp, br_hmp, tl_emb, br_emb, tl_off, br_off, targets)[source]

Compute losses for single level.

Parameters:
  • tl_hmp (Tensor) – Top-left corner heatmap for current level with shape (N, num_classes, H, W).
  • br_hmp (Tensor) – Bottom-right corner heatmap for current level with shape (N, num_classes, H, W).
  • tl_emb (Tensor) – Top-left corner embedding for current level with shape (N, corner_emb_channels, H, W).
  • br_emb (Tensor) – Bottom-right corner embedding for current level with shape (N, corner_emb_channels, H, W).
  • tl_off (Tensor) – Top-left corner offset for current level with shape (N, corner_offset_channels, H, W).
  • br_off (Tensor) – Bottom-right corner offset for current level with shape (N, corner_offset_channels, H, W).
  • targets (dict) – Corner target generated by get_targets.
Returns:

Losses of the head’s differnet branches containing the following losses:

  • det_loss (Tensor): Corner keypoint loss.
  • pull_loss (Tensor): Part one of AssociativeEmbedding loss.
  • push_loss (Tensor): Part two of AssociativeEmbedding loss.
  • off_loss (Tensor): Corner offset loss.

Return type:

tuple[torch.Tensor]

class mmdet.models.dense_heads.YOLACTHead(num_classes, in_channels, anchor_generator={'octave_base_scale': 3, 'ratios': [0.5, 1.0, 2.0], 'scales_per_octave': 1, 'strides': [8, 16, 32, 64, 128], 'type': 'AnchorGenerator'}, loss_cls={'loss_weight': 1.0, 'reduction': 'none', 'type': 'CrossEntropyLoss', 'use_sigmoid': False}, loss_bbox={'beta': 1.0, 'loss_weight': 1.5, 'type': 'SmoothL1Loss'}, num_head_convs=1, num_protos=32, use_ohem=True, conv_cfg=None, norm_cfg=None, init_cfg={'bias': 0, 'distribution': 'uniform', 'layer': 'Conv2d', 'type': 'Xavier'}, **kwargs)[source]

YOLACT box head used in https://arxiv.org/abs/1904.02689.

Note that YOLACT head is a light version of RetinaNet head. Four differences are described as follows:

  1. YOLACT box head has three-times fewer anchors.
  2. YOLACT box head shares the convs for box and cls branches.
  3. YOLACT box head uses OHEM instead of Focal loss.
  4. YOLACT box head predicts a set of mask coefficients for each box.
Parameters:
  • num_classes (int) – Number of categories excluding the background category.
  • in_channels (int) – Number of channels in the input feature map.
  • anchor_generator (dict) – Config dict for anchor generator
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox (dict) – Config of localization loss.
  • num_head_convs (int) – Number of the conv layers shared by box and cls branches.
  • num_protos (int) – Number of the mask coefficients.
  • use_ohem (bool) – If true, loss_single_OHEM will be used for cls loss calculation. If false, loss_single will be used.
  • conv_cfg (dict) – Dictionary to construct and config conv layer.
  • norm_cfg (dict) – Dictionary to construct and config norm layer.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward_single(x)[source]

Forward feature of a single scale level.

Parameters:x (Tensor) – Features of a single scale level.
Returns:cls_score (Tensor): Cls scores for a single scale level the channels number is num_anchors * num_classes. bbox_pred (Tensor): Box energies / deltas for a single scale level, the channels number is num_anchors * 4. coeff_pred (Tensor): Mask coefficients for a single scale level, the channels number is num_anchors * num_protos.
Return type:tuple
get_bboxes(cls_scores, bbox_preds, coeff_preds, img_metas, cfg=None, rescale=False)[source]

“Similiar to func:AnchorHead.get_bboxes, but additionally processes coeff_preds.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level with shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • coeff_preds (list[Tensor]) – Mask coefficients for each scale level with shape (N, num_anchors * num_protos, H, W)
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used
  • rescale (bool) – If True, return boxes in original image space. Default: False.
Returns:

Each item in result_list is

a 3-tuple. The first item is an (n, 5) tensor, where the first 4 columns are bounding box positions (tl_x, tl_y, br_x, br_y) and the 5-th column is a score between 0 and 1. The second item is an (n,) tensor where each item is the predicted class label of the corresponding box. The third item is an (n, num_protos) tensor where each item is the predicted mask coefficients of instance inside the corresponding box.

Return type:

list[tuple[Tensor, Tensor, Tensor]]

loss(cls_scores, bbox_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

A combination of the func:AnchorHead.loss and func:SSDHead.loss.

When self.use_ohem == True, it functions like SSDHead.loss, otherwise, it follows AnchorHead.loss. Besides, it additionally returns sampling_results.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – Class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – Specify which bounding boxes can be ignored when computing the loss. Default: None
Returns:

dict[str, Tensor]: A dictionary of loss components. List[:obj:SamplingResult]: Sampler results for each image.

Return type:

tuple

loss_single_OHEM(cls_score, bbox_pred, anchors, labels, label_weights, bbox_targets, bbox_weights, num_total_samples)[source]

“See func:SSDHead.loss.

class mmdet.models.dense_heads.YOLACTSegmHead(num_classes, in_channels=256, loss_segm={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, init_cfg={'distribution': 'uniform', 'override': {'name': 'segm_conv'}, 'type': 'Xavier'})[source]

YOLACT segmentation head used in https://arxiv.org/abs/1904.02689.

Apply a semantic segmentation loss on feature space using layers that are only evaluated during training to increase performance with no speed penalty.

Parameters:
  • in_channels (int) – Number of channels in the input feature map.
  • num_classes (int) – Number of categories excluding the background category.
  • loss_segm (dict) – Config of semantic segmentation loss.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(x)[source]

Forward feature from the upstream network.

Parameters:x (Tensor) – Feature from the upstream network, which is a 4D-tensor.
Returns:
Predicted semantic segmentation map with shape
(N, num_classes, H, W).
Return type:Tensor
get_targets(segm_pred, gt_masks, gt_labels)[source]

Compute semantic segmentation targets for each image.

Parameters:
  • segm_pred (Tensor) – Predicted semantic segmentation map with shape (num_classes, H, W).
  • gt_masks (Tensor) – Ground truth masks for each image with the same shape of the input image.
  • gt_labels (Tensor) – Class indices corresponding to each box.
Returns:

Semantic segmentation targets with shape

(num_classes, H, W).

Return type:

Tensor

loss(segm_pred, gt_masks, gt_labels)[source]

Compute loss of the head.

Parameters:
  • segm_pred (list[Tensor]) – Predicted semantic segmentation map with shape (N, num_classes, H, W).
  • gt_masks (list[Tensor]) – Ground truth masks for each image with the same shape of the input image.
  • gt_labels (list[Tensor]) – Class indices corresponding to each box.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

simple_test(feats, img_metas, rescale=False)[source]

Test function without test-time augmentation.

class mmdet.models.dense_heads.YOLACTProtonet(num_classes, in_channels=256, proto_channels=(256, 256, 256, None, 256, 32), proto_kernel_sizes=(3, 3, 3, -2, 3, 1), include_last_relu=True, num_protos=32, loss_mask_weight=1.0, max_masks_to_train=100, init_cfg={'distribution': 'uniform', 'override': {'name': 'protonet'}, 'type': 'Xavier'})[source]

YOLACT mask head used in https://arxiv.org/abs/1904.02689.

This head outputs the mask prototypes for YOLACT.

Parameters:
  • in_channels (int) – Number of channels in the input feature map.
  • proto_channels (tuple[int]) – Output channels of protonet convs.
  • proto_kernel_sizes (tuple[int]) – Kernel sizes of protonet convs.
  • include_last_relu (Bool) – If keep the last relu of protonet.
  • num_protos (int) – Number of prototypes.
  • num_classes (int) – Number of categories excluding the background category.
  • loss_mask_weight (float) – Reweight the mask loss by this factor.
  • max_masks_to_train (int) – Maximum number of masks to train for each image.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
crop(masks, boxes, padding=1)[source]

Crop predicted masks by zeroing out everything not in the predicted bbox.

Parameters:
  • masks (Tensor) – shape [H, W, N].
  • boxes (Tensor) – bbox coords in relative point form with shape [N, 4].
Returns:

The cropped masks.

Return type:

Tensor

forward(x, coeff_pred, bboxes, img_meta, sampling_results=None)[source]

Forward feature from the upstream network to get prototypes and linearly combine the prototypes, using masks coefficients, into instance masks. Finally, crop the instance masks with given bboxes.

Parameters:
  • x (Tensor) – Feature from the upstream network, which is a 4D-tensor.
  • coeff_pred (list[Tensor]) – Mask coefficients for each scale level with shape (N, num_anchors * num_protos, H, W).
  • bboxes (list[Tensor]) – Box used for cropping with shape (N, num_anchors * 4, H, W). During training, they are ground truth boxes. During testing, they are predicted boxes.
  • img_meta (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • sampling_results (List[:obj:SamplingResult]) – Sampler results for each image.
Returns:

Predicted instance segmentation masks.

Return type:

list[Tensor]

get_seg_masks(mask_pred, label_pred, img_meta, rescale)[source]

Resize, binarize, and format the instance mask predictions.

Parameters:
  • mask_pred (Tensor) – shape (N, H, W).
  • label_pred (Tensor) – shape (N, ).
  • img_meta (dict) – Meta information of each image, e.g., image size, scaling factor, etc.
  • rescale (bool) – If rescale is False, then returned masks will fit the scale of imgs[0].
Returns:

Mask predictions grouped by their predicted classes.

Return type:

list[ndarray]

get_targets(mask_pred, gt_masks, pos_assigned_gt_inds)[source]

Compute instance segmentation targets for each image.

Parameters:
  • mask_pred (Tensor) – Predicted prototypes with shape (num_classes, H, W).
  • gt_masks (Tensor) – Ground truth masks for each image with the same shape of the input image.
  • pos_assigned_gt_inds (Tensor) – GT indices of the corresponding positive samples.
Returns:

Instance segmentation targets with shape

(num_instances, H, W).

Return type:

Tensor

loss(mask_pred, gt_masks, gt_bboxes, img_meta, sampling_results)[source]

Compute loss of the head.

Parameters:
  • mask_pred (list[Tensor]) – Predicted prototypes with shape (num_classes, H, W).
  • gt_masks (list[Tensor]) – Ground truth masks for each image with the same shape of the input image.
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • img_meta (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • sampling_results (List[:obj:SamplingResult]) – Sampler results for each image.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

sanitize_coordinates(x1, x2, img_size, padding=0, cast=True)[source]

Sanitizes the input coordinates so that x1 < x2, x1 != x2, x1 >= 0, and x2 <= image_size. Also converts from relative to absolute coordinates and casts the results to long tensors.

Warning: this does things in-place behind the scenes so copy if necessary.

Parameters:
  • _x1 (Tensor) – shape (N, ).
  • _x2 (Tensor) – shape (N, ).
  • img_size (int) – Size of the input image.
  • padding (int) – x1 >= padding, x2 <= image_size-padding.
  • cast (bool) – If cast is false, the result won’t be cast to longs.
Returns:

x1 (Tensor): Sanitized _x1. x2 (Tensor): Sanitized _x2.

Return type:

tuple

simple_test(feats, det_bboxes, det_labels, det_coeffs, img_metas, rescale=False)[source]

Test function without test-time augmentation.

Parameters:
  • feats (tuple[torch.Tensor]) – Multi-level features from the upstream network, each is a 4D-tensor.
  • det_bboxes (list[Tensor]) – BBox results of each image. each element is (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1.
  • det_labels (list[Tensor]) – BBox results of each image. each element is (n, ) tensor, each element represents the class label of the corresponding box.
  • det_coeffs (list[Tensor]) – BBox coefficient of each image. each element is (n, m) tensor, m is vector length.
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

encoded masks. The c-th item in the outer list

corresponds to the c-th class. Given the c-th outer list, the i-th item in that inner list is the mask for the i-th box with class label c.

Return type:

list[list]

class mmdet.models.dense_heads.YOLOV3Head(num_classes, in_channels, out_channels=(1024, 512, 256), anchor_generator={'base_sizes': [[(116, 90), (156, 198), (373, 326)], [(30, 61), (62, 45), (59, 119)], [(10, 13), (16, 30), (33, 23)]], 'strides': [32, 16, 8], 'type': 'YOLOAnchorGenerator'}, bbox_coder={'type': 'YOLOBBoxCoder'}, featmap_strides=[32, 16, 8], one_hot_smoother=0.0, conv_cfg=None, norm_cfg={'requires_grad': True, 'type': 'BN'}, act_cfg={'negative_slope': 0.1, 'type': 'LeakyReLU'}, loss_cls={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_conf={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_xy={'loss_weight': 1.0, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_wh={'loss_weight': 1.0, 'type': 'MSELoss'}, train_cfg=None, test_cfg=None, init_cfg={'override': {'name': 'convs_pred'}, 'std': 0.01, 'type': 'Normal'})[source]

YOLOV3Head Paper link: https://arxiv.org/abs/1804.02767.

Parameters:
  • num_classes (int) – The number of object classes (w/o background)
  • in_channels (List[int]) – Number of input channels per scale.
  • out_channels (List[int]) – The number of output channels per scale before the final 1x1 layer. Default: (1024, 512, 256).
  • anchor_generator (dict) – Config dict for anchor generator
  • bbox_coder (dict) – Config of bounding box coder.
  • featmap_strides (List[int]) – The stride of each scale. Should be in descending order. Default: (32, 16, 8).
  • one_hot_smoother (float) – Set a non-zero value to enable label-smooth Default: 0.
  • conv_cfg (dict) – Config dict for convolution layer. Default: None.
  • norm_cfg (dict) – Dictionary to construct and config norm layer. Default: dict(type=’BN’, requires_grad=True)
  • act_cfg (dict) – Config dict for activation layer. Default: dict(type=’LeakyReLU’, negative_slope=0.1).
  • loss_cls (dict) – Config of classification loss.
  • loss_conf (dict) – Config of confidence loss.
  • loss_xy (dict) – Config of xy coordinate loss.
  • loss_wh (dict) – Config of wh coordinate loss.
  • train_cfg (dict) – Training config of YOLOV3 head. Default: None.
  • test_cfg (dict) – Testing config of YOLOV3 head. Default: None.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
aug_test(feats, img_metas, rescale=False)[source]

Test function with test time augmentation.

Parameters:
  • feats (list[Tensor]) – the outer list indicates test-time augmentations and inner Tensor should have a shape NxCxHxW, which contains features for all images in the batch.
  • img_metas (list[list[dict]]) – the outer list indicates test-time augs (multiscale, flip, etc.) and the inner list indicates images in a batch. each dict has image information.
  • rescale (bool, optional) – Whether to rescale the results. Defaults to False.
Returns:

bbox results of each class

Return type:

list[ndarray]

forward(feats)[source]

Forward features from the upstream network.

Parameters:feats (tuple[Tensor]) – Features from the upstream network, each is a 4D-tensor.
Returns:
A tuple of multi-level predication map, each is a
4D-tensor of shape (batch_size, 5+num_classes, height, width).
Return type:tuple[Tensor]
get_bboxes(pred_maps, img_metas, cfg=None, rescale=False, with_nms=True)[source]

Transform network output for a batch into bbox predictions.

Parameters:
  • pred_maps (list[Tensor]) – Raw predictions for a batch of images.
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • cfg (mmcv.Config | None) – Test / postprocessing configuration, if None, test_cfg would be used. Default: None.
  • rescale (bool) – If True, return boxes in original image space. Default: False.
  • with_nms (bool) – If True, do nms before return boxes. Default: True.
Returns:

Each item in result_list is 2-tuple.

The first item is an (n, 5) tensor, where 5 represent (tl_x, tl_y, br_x, br_y, score) and the score between 0 and 1. The shape of the second tensor in the tuple is (n,), and each element represents the class label of the corresponding box.

Return type:

list[tuple[Tensor, Tensor]]

get_targets(anchor_list, responsible_flag_list, gt_bboxes_list, gt_labels_list)[source]

Compute target maps for anchors in multiple images.

Parameters:
  • anchor_list (list[list[Tensor]]) – Multi level anchors of each image. The outer list indicates images, and the inner list corresponds to feature levels of the image. Each element of the inner list is a tensor of shape (num_total_anchors, 4).
  • responsible_flag_list (list[list[Tensor]]) – Multi level responsible flags of each image. Each element is a tensor of shape (num_total_anchors, )
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box.
Returns:

Usually returns a tuple containing learning targets.
  • target_map_list (list[Tensor]): Target map of each level.
  • neg_map_list (list[Tensor]): Negative map of each level.

Return type:

tuple

init_weights()[source]

Initialize the weights.

loss(pred_maps, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute loss of the head.

Parameters:
  • pred_maps (list[Tensor]) – Prediction map for each scale level, shape (N, num_anchors * num_attrib, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (None | list[Tensor]) – specify which bounding boxes can be ignored when computing the loss.
Returns:

A dictionary of loss components.

Return type:

dict[str, Tensor]

loss_single(pred_map, target_map, neg_map)[source]

Compute loss of a single image from a batch.

Parameters:
  • pred_map (Tensor) – Raw predictions for a single level.
  • target_map (Tensor) – The Ground-Truth target for a single level.
  • neg_map (Tensor) – The negative masks for a single level.
Returns:

loss_cls (Tensor): Classification loss. loss_conf (Tensor): Confidence loss. loss_xy (Tensor): Regression loss of x, y coordinate. loss_wh (Tensor): Regression loss of w, h coordinate.

Return type:

tuple

num_attrib

number of attributes in pred_map, bboxes (4) + objectness (1) + num_classes

Type:int
class mmdet.models.dense_heads.PAAHead(*args, topk=9, score_voting=True, covariance_type='diag', **kwargs)[source]

Head of PAAAssignment: Probabilistic Anchor Assignment with IoU Prediction for Object Detection.

Code is modified from the official github repo.

More details can be found in the paper .

Parameters:
  • topk (int) – Select topk samples with smallest loss in each level.
  • score_voting (bool) – Whether to use score voting in post-process.
  • covariance_type

    String describing the type of covariance parameters to be used in sklearn.mixture.GaussianMixture. It must be one of:

    • ’full’: each component has its own general covariance matrix
    • ’tied’: all components share the same general covariance matrix
    • ’diag’: each component has its own diagonal covariance matrix
    • ’spherical’: each component has its own single variance

    Default: ‘diag’. From ‘full’ to ‘spherical’, the gmm fitting process is faster yet the performance could be influenced. For most cases, ‘diag’ should be a good choice.

get_pos_loss(anchors, cls_score, bbox_pred, label, label_weight, bbox_target, bbox_weight, pos_inds)[source]

Calculate loss of all potential positive samples obtained from first match process.

Parameters:
  • anchors (list[Tensor]) – Anchors of each scale.
  • cls_score (Tensor) – Box scores of single image with shape (num_anchors, num_classes)
  • bbox_pred (Tensor) – Box energies / deltas of single image with shape (num_anchors, 4)
  • label (Tensor) – classification target of each anchor with shape (num_anchors,)
  • label_weight (Tensor) – Classification loss weight of each anchor with shape (num_anchors).
  • bbox_target (dict) – Regression target of each anchor with shape (num_anchors, 4).
  • bbox_weight (Tensor) – Bbox weight of each anchor with shape (num_anchors, 4).
  • pos_inds (Tensor) – Index of all positive samples got from first assign process.
Returns:

Losses of all positive samples in single image.

Return type:

Tensor

get_targets(anchor_list, valid_flag_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, label_channels=1, unmap_outputs=True)[source]

Get targets for PAA head.

This method is almost the same as AnchorHead.get_targets(). We direct return the results from _get_targets_single instead map it to levels by images_to_levels function.

Parameters:
  • anchor_list (list[list[Tensor]]) – Multi level anchors of each image. The outer list indicates images, and the inner list corresponds to feature levels of the image. Each element of the inner list is a tensor of shape (num_anchors, 4).
  • valid_flag_list (list[list[Tensor]]) – Multi level valid flags of each image. The outer list indicates images, and the inner list corresponds to feature levels of the image. Each element of the inner list is a tensor of shape (num_anchors, )
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • img_metas (list[dict]) – Meta info of each image.
  • gt_bboxes_ignore_list (list[Tensor]) – Ground truth bboxes to be ignored.
  • gt_labels_list (list[Tensor]) – Ground truth labels of each box.
  • label_channels (int) – Channel of label.
  • unmap_outputs (bool) – Whether to map outputs back to the original set of anchors.
Returns:

Usually returns a tuple containing learning targets.

  • labels (list[Tensor]): Labels of all anchors, each with
    shape (num_anchors,).
  • label_weights (list[Tensor]): Label weights of all anchor.
    each with shape (num_anchors,).
  • bbox_targets (list[Tensor]): BBox targets of all anchors.
    each with shape (num_anchors, 4).
  • bbox_weights (list[Tensor]): BBox weights of all anchors.
    each with shape (num_anchors, 4).
  • pos_inds (list[Tensor]): Contains all index of positive
    sample in all anchor.
  • gt_inds (list[Tensor]): Contains all gt_index of positive
    sample in all anchor.

Return type:

tuple

gmm_separation_scheme(gmm_assignment, scores, pos_inds_gmm)[source]

A general separation scheme for gmm model.

It separates a GMM distribution of candidate samples into three parts, 0 1 and uncertain areas, and you can implement other separation schemes by rewriting this function.

Parameters:
  • gmm_assignment (Tensor) – The prediction of GMM which is of shape (num_samples,). The 0/1 value indicates the distribution that each sample comes from.
  • scores (Tensor) – The probability of sample coming from the fit GMM distribution. The tensor is of shape (num_samples,).
  • pos_inds_gmm (Tensor) – All the indexes of samples which are used to fit GMM model. The tensor is of shape (num_samples,)
Returns:

The indices of positive and ignored samples.

  • pos_inds_temp (Tensor): Indices of positive samples.
  • ignore_inds_temp (Tensor): Indices of ignore samples.

Return type:

tuple[Tensor]

loss(cls_scores, bbox_preds, iou_preds, gt_bboxes, gt_labels, img_metas, gt_bboxes_ignore=None)[source]

Compute losses of the head.

Parameters:
  • cls_scores (list[Tensor]) – Box scores for each scale level Has shape (N, num_anchors * num_classes, H, W)
  • bbox_preds (list[Tensor]) – Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W)
  • iou_preds (list[Tensor]) – iou_preds for each scale level with shape (N, num_anchors * 1, H, W)
  • gt_bboxes (list[Tensor]) – Ground truth bboxes for each image with shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
  • gt_labels (list[Tensor]) – class indices corresponding to each box
  • img_metas (list[dict]) – Meta information of each image, e.g., image size, scaling factor, etc.
  • gt_bboxes_ignore (list[Tensor] | None) – Specify which bounding boxes can be ignored when are computing the loss.
Returns:

A dictionary of loss gmm_assignment.

Return type:

dict[str, Tensor]

paa_reassign(pos_losses, label, label_weight, bbox_weight, pos_inds, pos_gt_inds, anchors)[source]

Fit loss to GMM distribution and separate positive, ignore, negative samples again with GMM model.

Parameters:
  • pos_losses (Tensor) – Losses of all positive samples in single image.
  • label (Tensor) – classification target of each anchor with shape (num_anchors,)
  • label_weight (Tensor) – Classification loss weight of each anchor with shape (num_anchors).
  • bbox_weight (Tensor) – Bbox weight of each anchor with shape (num_anchors, 4).
  • pos_inds (Tensor) – Index of all positive samples got from first assign process.
  • pos_gt_inds (Tensor) – Gt_index of all positive samples got from first assign process.
  • anchors (list[Tensor]) – Anchors of each scale.
Returns:

Usually returns a tuple containing learning targets.

  • label (Tensor): classification target of each anchor after paa assign, with shape (num_anchors,)
  • label_weight (Tensor): Classification loss weight of each anchor after paa assign, with shape (num_anchors).
  • bbox_weight (Tensor): Bbox weight of each anchor with shape (num_anchors, 4).
  • num_pos (int): The number of positive samples after paa assign.

Return type:

tuple

score_voting(det_bboxes, det_labels, mlvl_bboxes, mlvl_nms_scores, score_thr)[source]

Implementation of score voting method works on each remaining boxes after NMS procedure.

Parameters:
  • det_bboxes (Tensor) – Remaining boxes after NMS procedure, with shape (k, 5), each dimension means (x1, y1, x2, y2, score).
  • det_labels (Tensor) – The label of remaining boxes, with shape (k, 1),Labels are 0-based.
  • mlvl_bboxes (Tensor) – All boxes before the NMS procedure, with shape (num_anchors,4).
  • mlvl_nms_scores (Tensor) – The scores of all boxes which is used in the NMS procedure, with shape (num_anchors, num_class)
  • mlvl_iou_preds (Tensor) – The predictions of IOU of all boxes before the NMS procedure, with shape (num_anchors, 1)
  • score_thr (float) – The score threshold of bboxes.
Returns:

Usually returns a tuple containing voting results.

  • det_bboxes_voted (Tensor): Remaining boxes after
    score voting procedure, with shape (k, 5), each dimension means (x1, y1, x2, y2, score).
  • det_labels_voted (Tensor): Label of remaining bboxes
    after voting, with shape (num_anchors,).

Return type:

tuple

class mmdet.models.dense_heads.SABLRetinaHead(num_classes, in_channels, stacked_convs=4, feat_channels=256, approx_anchor_generator={'octave_base_scale': 4, 'ratios': [0.5, 1.0, 2.0], 'scales_per_octave': 3, 'strides': [8, 16, 32, 64, 128], 'type': 'AnchorGenerator'}, square_anchor_generator={'ratios': [1.0], 'scales': [4], 'strides': [8, 16, 32, 64, 128], 'type': 'AnchorGenerator'}, conv_cfg=None, norm_cfg=None, bbox_coder={'num_buckets': 14, 'scale_factor': 3.0, 'type': 'BucketingBBoxCoder'}, reg_decoded_bbox=False, train_cfg=None, test_cfg=None, loss_cls={'alpha': 0.25, 'gamma': 2.0, 'loss_weight': 1.0, 'type': 'FocalLoss', 'use_sigmoid': True}, loss_bbox_cls={'loss_weight': 1.5, 'type': 'CrossEntropyLoss', 'use_sigmoid': True}, loss_bbox_reg={'beta': 0.1111111111111111, 'loss_weight': 1.5, 'type': 'SmoothL1Loss'}, init_cfg={'layer': 'Conv2d', 'override': {'bias_prob': 0.01, 'name': 'retina_cls', 'std': 0.01, 'type': 'Normal'}, 'std': 0.01, 'type': 'Normal'})[source]

Side-Aware Boundary Localization (SABL) for RetinaNet.

The anchor generation, assigning and sampling in SABLRetinaHead are the same as GuidedAnchorHead for guided anchoring.

Please refer to https://arxiv.org/abs/1912.04260 for more details.

Parameters:
  • num_classes (int) – Number of classes.
  • in_channels (int) – Number of channels in the input feature map.
  • stacked_convs (int) – Number of Convs for classification and regression branches. Defaults to 4.
  • feat_channels (int) – Number of hidden channels. Defaults to 256.
  • approx_anchor_generator (dict) – Config dict for approx generator.
  • square_anchor_generator (dict) – Config dict for square generator.
  • conv_cfg (dict) – Config dict for ConvModule. Defaults to None.
  • norm_cfg (dict) – Config dict for Norm Layer. Defaults to None.
  • bbox_coder (dict) – Config dict for bbox coder.
  • reg_decoded_bbox (bool) – If true, the regression loss would be applied directly on decoded bounding boxes, converting both the predicted boxes and regression targets to absolute coordinates format. Default False. It should be True when using IoULoss, GIoULoss, or DIoULoss in the bbox head.
  • train_cfg (dict) – Training config of SABLRetinaHead.
  • test_cfg (dict) – Testing config of SABLRetinaHead.
  • loss_cls (dict) – Config of classification loss.
  • loss_bbox_cls (dict) – Config of classification loss for bbox branch.
  • loss_bbox_reg (dict) – Config of regression loss for bbox branch.
  • init_cfg (dict or list[dict], optional) – Initialization config dict.
forward(feats)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

get_anchors(featmap_sizes, img_metas, device='cuda')[source]

Get squares according to feature map sizes and guided anchors.

Parameters:
  • featmap_sizes (list[tuple]) – Multi-level feature map sizes.
  • img_metas (list[dict]) – Image meta info.
  • device (torch.device | str) – device for returned tensors
Returns:

square approxs of each image

Return type:

tuple

get_bboxes(cls_scores, bbox_preds, img_metas, cfg=None, rescale=False)[source]

Transform network output for a batch into bbox predictions.

get_target(approx_list, inside_flag_list, square_list, gt_bboxes_list, img_metas, gt_bboxes_ignore_list=None, gt_labels_list=None, label_channels=None, sampling=True, unmap_outputs=True)[source]

Compute bucketing targets. :param approx_list: Multi level approxs of each image. :type approx_list: list[list] :param inside_flag_list: Multi level inside flags of each

image.
Parameters:
  • square_list (list[list]) – Multi level squares of each image.
  • gt_bboxes_list (list[Tensor]) – Ground truth bboxes of each image.
  • img_metas (list[dict]) – Meta info of each image.
  • gt_bboxes_ignore_list (list[Tensor]) – ignore list of gt bboxes.
  • gt_bboxes_list – Gt bboxes of each image.
  • label_channels (int) – Channel of label.
  • sampling (bool) – Sample Anchors or not.
  • unmap_outputs (bool) – unmap outputs or not.