Shortcuts

mmdet.core.evaluation.mean_ap 源代码

# Copyright (c) OpenMMLab. All rights reserved.
from multiprocessing import Pool

import mmcv
import numpy as np
from mmcv.utils import print_log
from terminaltables import AsciiTable

from .bbox_overlaps import bbox_overlaps
from .class_names import get_classes


[文档]def average_precision(recalls, precisions, mode='area'): """Calculate average precision (for single or multiple scales). Args: 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: float or ndarray: calculated average precision """ no_scale = False if recalls.ndim == 1: no_scale = True recalls = recalls[np.newaxis, :] precisions = precisions[np.newaxis, :] assert recalls.shape == precisions.shape and recalls.ndim == 2 num_scales = recalls.shape[0] ap = np.zeros(num_scales, dtype=np.float32) if mode == 'area': zeros = np.zeros((num_scales, 1), dtype=recalls.dtype) ones = np.ones((num_scales, 1), dtype=recalls.dtype) mrec = np.hstack((zeros, recalls, ones)) mpre = np.hstack((zeros, precisions, zeros)) for i in range(mpre.shape[1] - 1, 0, -1): mpre[:, i - 1] = np.maximum(mpre[:, i - 1], mpre[:, i]) for i in range(num_scales): ind = np.where(mrec[i, 1:] != mrec[i, :-1])[0] ap[i] = np.sum( (mrec[i, ind + 1] - mrec[i, ind]) * mpre[i, ind + 1]) elif mode == '11points': for i in range(num_scales): for thr in np.arange(0, 1 + 1e-3, 0.1): precs = precisions[i, recalls[i, :] >= thr] prec = precs.max() if precs.size > 0 else 0 ap[i] += prec ap /= 11 else: raise ValueError( 'Unrecognized mode, only "area" and "11points" are supported') if no_scale: ap = ap[0] return ap
def tpfp_imagenet(det_bboxes, gt_bboxes, gt_bboxes_ignore=None, default_iou_thr=0.5, area_ranges=None, use_legacy_coordinate=False): """Check if detected bboxes are true positive or false positive. Args: det_bbox (ndarray): Detected bboxes of this image, of shape (m, 5). gt_bboxes (ndarray): GT bboxes of this image, of shape (n, 4). gt_bboxes_ignore (ndarray): Ignored gt bboxes of this image, of shape (k, 4). Default: None default_iou_thr (float): IoU threshold to be considered as matched for medium and large bboxes (small ones have special rules). Default: 0.5. area_ranges (list[tuple] | None): Range of bbox areas to be evaluated, in the format [(min1, max1), (min2, max2), ...]. Default: None. use_legacy_coordinate (bool): Whether to use coordinate system in mmdet v1.x. which means width, height should be calculated as 'x2 - x1 + 1` and 'y2 - y1 + 1' respectively. Default: False. Returns: tuple[np.ndarray]: (tp, fp) whose elements are 0 and 1. The shape of each array is (num_scales, m). """ if not use_legacy_coordinate: extra_length = 0. else: extra_length = 1. # an indicator of ignored gts gt_ignore_inds = np.concatenate( (np.zeros(gt_bboxes.shape[0], dtype=np.bool), np.ones(gt_bboxes_ignore.shape[0], dtype=np.bool))) # stack gt_bboxes and gt_bboxes_ignore for convenience gt_bboxes = np.vstack((gt_bboxes, gt_bboxes_ignore)) num_dets = det_bboxes.shape[0] num_gts = gt_bboxes.shape[0] if area_ranges is None: area_ranges = [(None, None)] num_scales = len(area_ranges) # tp and fp are of shape (num_scales, num_gts), each row is tp or fp # of a certain scale. tp = np.zeros((num_scales, num_dets), dtype=np.float32) fp = np.zeros((num_scales, num_dets), dtype=np.float32) if gt_bboxes.shape[0] == 0: if area_ranges == [(None, None)]: fp[...] = 1 else: det_areas = ( det_bboxes[:, 2] - det_bboxes[:, 0] + extra_length) * ( det_bboxes[:, 3] - det_bboxes[:, 1] + extra_length) for i, (min_area, max_area) in enumerate(area_ranges): fp[i, (det_areas >= min_area) & (det_areas < max_area)] = 1 return tp, fp ious = bbox_overlaps( det_bboxes, gt_bboxes - 1, use_legacy_coordinate=use_legacy_coordinate) gt_w = gt_bboxes[:, 2] - gt_bboxes[:, 0] + extra_length gt_h = gt_bboxes[:, 3] - gt_bboxes[:, 1] + extra_length iou_thrs = np.minimum((gt_w * gt_h) / ((gt_w + 10.0) * (gt_h + 10.0)), default_iou_thr) # sort all detections by scores in descending order sort_inds = np.argsort(-det_bboxes[:, -1]) for k, (min_area, max_area) in enumerate(area_ranges): gt_covered = np.zeros(num_gts, dtype=bool) # if no area range is specified, gt_area_ignore is all False if min_area is None: gt_area_ignore = np.zeros_like(gt_ignore_inds, dtype=bool) else: gt_areas = gt_w * gt_h gt_area_ignore = (gt_areas < min_area) | (gt_areas >= max_area) for i in sort_inds: max_iou = -1 matched_gt = -1 # find best overlapped available gt for j in range(num_gts): # different from PASCAL VOC: allow finding other gts if the # best overlapped ones are already matched by other det bboxes if gt_covered[j]: continue elif ious[i, j] >= iou_thrs[j] and ious[i, j] > max_iou: max_iou = ious[i, j] matched_gt = j # there are 4 cases for a det bbox: # 1. it matches a gt, tp = 1, fp = 0 # 2. it matches an ignored gt, tp = 0, fp = 0 # 3. it matches no gt and within area range, tp = 0, fp = 1 # 4. it matches no gt but is beyond area range, tp = 0, fp = 0 if matched_gt >= 0: gt_covered[matched_gt] = 1 if not (gt_ignore_inds[matched_gt] or gt_area_ignore[matched_gt]): tp[k, i] = 1 elif min_area is None: fp[k, i] = 1 else: bbox = det_bboxes[i, :4] area = (bbox[2] - bbox[0] + extra_length) * ( bbox[3] - bbox[1] + extra_length) if area >= min_area and area < max_area: fp[k, i] = 1 return tp, fp def tpfp_default(det_bboxes, gt_bboxes, gt_bboxes_ignore=None, iou_thr=0.5, area_ranges=None, use_legacy_coordinate=False): """Check if detected bboxes are true positive or false positive. Args: det_bbox (ndarray): Detected bboxes of this image, of shape (m, 5). gt_bboxes (ndarray): GT bboxes of this image, of shape (n, 4). gt_bboxes_ignore (ndarray): Ignored gt bboxes of this image, of shape (k, 4). Default: None iou_thr (float): IoU threshold to be considered as matched. Default: 0.5. area_ranges (list[tuple] | None): Range of bbox areas to be evaluated, in the format [(min1, max1), (min2, max2), ...]. Default: None. use_legacy_coordinate (bool): Whether to use coordinate system in mmdet v1.x. which means width, height should be calculated as 'x2 - x1 + 1` and 'y2 - y1 + 1' respectively. Default: False. Returns: tuple[np.ndarray]: (tp, fp) whose elements are 0 and 1. The shape of each array is (num_scales, m). """ if not use_legacy_coordinate: extra_length = 0. else: extra_length = 1. # an indicator of ignored gts gt_ignore_inds = np.concatenate( (np.zeros(gt_bboxes.shape[0], dtype=np.bool), np.ones(gt_bboxes_ignore.shape[0], dtype=np.bool))) # stack gt_bboxes and gt_bboxes_ignore for convenience gt_bboxes = np.vstack((gt_bboxes, gt_bboxes_ignore)) num_dets = det_bboxes.shape[0] num_gts = gt_bboxes.shape[0] if area_ranges is None: area_ranges = [(None, None)] num_scales = len(area_ranges) # tp and fp are of shape (num_scales, num_gts), each row is tp or fp of # a certain scale tp = np.zeros((num_scales, num_dets), dtype=np.float32) fp = np.zeros((num_scales, num_dets), dtype=np.float32) # if there is no gt bboxes in this image, then all det bboxes # within area range are false positives if gt_bboxes.shape[0] == 0: if area_ranges == [(None, None)]: fp[...] = 1 else: det_areas = ( det_bboxes[:, 2] - det_bboxes[:, 0] + extra_length) * ( det_bboxes[:, 3] - det_bboxes[:, 1] + extra_length) for i, (min_area, max_area) in enumerate(area_ranges): fp[i, (det_areas >= min_area) & (det_areas < max_area)] = 1 return tp, fp ious = bbox_overlaps( det_bboxes, gt_bboxes, use_legacy_coordinate=use_legacy_coordinate) # for each det, the max iou with all gts ious_max = ious.max(axis=1) # for each det, which gt overlaps most with it ious_argmax = ious.argmax(axis=1) # sort all dets in descending order by scores sort_inds = np.argsort(-det_bboxes[:, -1]) for k, (min_area, max_area) in enumerate(area_ranges): gt_covered = np.zeros(num_gts, dtype=bool) # if no area range is specified, gt_area_ignore is all False if min_area is None: gt_area_ignore = np.zeros_like(gt_ignore_inds, dtype=bool) else: gt_areas = (gt_bboxes[:, 2] - gt_bboxes[:, 0] + extra_length) * ( gt_bboxes[:, 3] - gt_bboxes[:, 1] + extra_length) gt_area_ignore = (gt_areas < min_area) | (gt_areas >= max_area) for i in sort_inds: if ious_max[i] >= iou_thr: matched_gt = ious_argmax[i] if not (gt_ignore_inds[matched_gt] or gt_area_ignore[matched_gt]): if not gt_covered[matched_gt]: gt_covered[matched_gt] = True tp[k, i] = 1 else: fp[k, i] = 1 # otherwise ignore this detected bbox, tp = 0, fp = 0 elif min_area is None: fp[k, i] = 1 else: bbox = det_bboxes[i, :4] area = (bbox[2] - bbox[0] + extra_length) * ( bbox[3] - bbox[1] + extra_length) if area >= min_area and area < max_area: fp[k, i] = 1 return tp, fp def get_cls_results(det_results, annotations, class_id): """Get det results and gt information of a certain class. Args: det_results (list[list]): Same as `eval_map()`. annotations (list[dict]): Same as `eval_map()`. class_id (int): ID of a specific class. Returns: tuple[list[np.ndarray]]: detected bboxes, gt bboxes, ignored gt bboxes """ cls_dets = [img_res[class_id] for img_res in det_results] cls_gts = [] cls_gts_ignore = [] for ann in annotations: gt_inds = ann['labels'] == class_id cls_gts.append(ann['bboxes'][gt_inds, :]) if ann.get('labels_ignore', None) is not None: ignore_inds = ann['labels_ignore'] == class_id cls_gts_ignore.append(ann['bboxes_ignore'][ignore_inds, :]) else: cls_gts_ignore.append(np.empty((0, 4), dtype=np.float32)) return cls_dets, cls_gts, cls_gts_ignore
[文档]def eval_map(det_results, annotations, scale_ranges=None, iou_thr=0.5, dataset=None, logger=None, tpfp_fn=None, nproc=4, use_legacy_coordinate=False): """Evaluate mAP of a dataset. Args: 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, :func:`tpfp_default` is used as default unless dataset is 'det' or 'vid' (:func:`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. use_legacy_coordinate (bool): Whether to use coordinate system in mmdet v1.x. which means width, height should be calculated as 'x2 - x1 + 1` and 'y2 - y1 + 1' respectively. Default: False. Returns: tuple: (mAP, [dict, dict, ...]) """ assert len(det_results) == len(annotations) if not use_legacy_coordinate: extra_length = 0. else: extra_length = 1. num_imgs = len(det_results) num_scales = len(scale_ranges) if scale_ranges is not None else 1 num_classes = len(det_results[0]) # positive class num area_ranges = ([(rg[0]**2, rg[1]**2) for rg in scale_ranges] if scale_ranges is not None else None) pool = Pool(nproc) eval_results = [] for i in range(num_classes): # get gt and det bboxes of this class cls_dets, cls_gts, cls_gts_ignore = get_cls_results( det_results, annotations, i) # choose proper function according to datasets to compute tp and fp if tpfp_fn is None: if dataset in ['det', 'vid']: tpfp_fn = tpfp_imagenet else: tpfp_fn = tpfp_default if not callable(tpfp_fn): raise ValueError( f'tpfp_fn has to be a function or None, but got {tpfp_fn}') # compute tp and fp for each image with multiple processes tpfp = pool.starmap( tpfp_fn, zip(cls_dets, cls_gts, cls_gts_ignore, [iou_thr for _ in range(num_imgs)], [area_ranges for _ in range(num_imgs)], [use_legacy_coordinate for _ in range(num_imgs)])) tp, fp = tuple(zip(*tpfp)) # calculate gt number of each scale # ignored gts or gts beyond the specific scale are not counted num_gts = np.zeros(num_scales, dtype=int) for j, bbox in enumerate(cls_gts): if area_ranges is None: num_gts[0] += bbox.shape[0] else: gt_areas = (bbox[:, 2] - bbox[:, 0] + extra_length) * ( bbox[:, 3] - bbox[:, 1] + extra_length) for k, (min_area, max_area) in enumerate(area_ranges): num_gts[k] += np.sum((gt_areas >= min_area) & (gt_areas < max_area)) # sort all det bboxes by score, also sort tp and fp cls_dets = np.vstack(cls_dets) num_dets = cls_dets.shape[0] sort_inds = np.argsort(-cls_dets[:, -1]) tp = np.hstack(tp)[:, sort_inds] fp = np.hstack(fp)[:, sort_inds] # calculate recall and precision with tp and fp tp = np.cumsum(tp, axis=1) fp = np.cumsum(fp, axis=1) eps = np.finfo(np.float32).eps recalls = tp / np.maximum(num_gts[:, np.newaxis], eps) precisions = tp / np.maximum((tp + fp), eps) # calculate AP if scale_ranges is None: recalls = recalls[0, :] precisions = precisions[0, :] num_gts = num_gts.item() mode = 'area' if dataset != 'voc07' else '11points' ap = average_precision(recalls, precisions, mode) eval_results.append({ 'num_gts': num_gts, 'num_dets': num_dets, 'recall': recalls, 'precision': precisions, 'ap': ap }) pool.close() if scale_ranges is not None: # shape (num_classes, num_scales) all_ap = np.vstack([cls_result['ap'] for cls_result in eval_results]) all_num_gts = np.vstack( [cls_result['num_gts'] for cls_result in eval_results]) mean_ap = [] for i in range(num_scales): if np.any(all_num_gts[:, i] > 0): mean_ap.append(all_ap[all_num_gts[:, i] > 0, i].mean()) else: mean_ap.append(0.0) else: aps = [] for cls_result in eval_results: if cls_result['num_gts'] > 0: aps.append(cls_result['ap']) mean_ap = np.array(aps).mean().item() if aps else 0.0 print_map_summary( mean_ap, eval_results, dataset, area_ranges, logger=logger) return mean_ap, eval_results
Read the Docs v: v2.17.0
Versions
latest
stable
v2.17.0
v2.16.0
v2.15.1
v2.15.0
Downloads
On Read the Docs
Project Home
Builds

Free document hosting provided by Read the Docs.