def _prepare_iou_dict(self):
pbar = tqdm(total=len(self.video_list))
for cnt, video in enumerate(self.video_list): # 对于每一个video
pbar.update(1)
fg = video.get_fg(self.fg_iou_thresh, self.gt_as_fg) # 前景
incomp, bg = video.get_negatives(
self.incomplete_iou_thresh, self.bg_iou_thresh,
self.bg_coverage_thresh, self.incomplete_overlap_thresh) #背景区域
self.prop_dict[video.id] = [fg, incomp,
bg] # 为每个video选择不同的前景 背景 和不完整片段
video_pool = fg + incomp + bg
# calculate act iou matrix
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
iou_array, overlap_array = segment_tiou(
prop_array,
prop_array) # 计算所有proposal之间的IOU and overlap 便于邻接矩阵 [N, N]
self.act_iou_dict[video.id] = iou_array
# calculate act distance matrix
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
distance_array = segment_distance(
prop_array, prop_array) # 计算proposal之间的距离 [N, N]
self.act_dis_dict[video.id] = distance_array
pbar.close()
def _prepare_iou_dict2(video_list, fg_iou_thresh, gt_as_fg,
incomplete_iou_thresh, bg_iou_thresh,
bg_coverage_thresh, incomplete_overlap_thresh):
prop_dict = {}
act_iou_dict = {}
act_dis_dict = {}
pbar = tqdm(total=len(video_list))
for cnt, video in enumerate(video_list):
pbar.update(1)
fg = video.get_fg(fg_iou_thresh, gt_as_fg)
incomp, bg = video.get_negatives(incomplete_iou_thresh, bg_iou_thresh,
bg_coverage_thresh,
incomplete_overlap_thresh)
prop_dict[video.id] = [fg, incomp, bg]
video_pool = fg + incomp + bg
# calculate act iou matrix
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
iou_array, overlap_array = segment_tiou(prop_array, prop_array)
act_iou_dict[video.id] = iou_array
# calculate act distance matrix
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
distance_array = segment_distance(prop_array, prop_array)
act_dis_dict[video.id] = distance_array
pbar.close()
return prop_dict, act_iou_dict, act_dis_dict
def _prepare_test_iou_dict(self):
pbar = tqdm(total=len(self.video_list))
for cnt, video in enumerate(self.video_list):
pbar.update(1)
video_pool = video.proposals
# calculate act iou matrix
prop_array = np.array([[prop.start_frame, prop.end_frame] for prop in video_pool])
iou_array, overlap_array = segment_tiou(prop_array, prop_array)
self.act_iou_dict[video.id] = iou_array
# calculate act distance matrix
prop_array = np.array([[prop.start_frame, prop.end_frame] for prop in video_pool])
distance_array = segment_distance(prop_array, prop_array)
self.act_dis_dict[video.id] = distance_array
pbar.close()
def _prepare_iou_dict(self):
pbar = tqdm(total=len(self.video_list))
for cnt, video in enumerate(self.video_list):
pbar.update(1)
fg = video.get_fg(self.fg_iou_thresh, self.gt_as_fg)
incomp, bg = video.get_negatives(self.incomplete_iou_thresh, self.bg_iou_thresh,
self.bg_coverage_thresh, self.incomplete_overlap_thresh)
self.prop_dict[video.id] = [fg, incomp, bg]
video_pool = fg + incomp + bg
# calculate act iou matrix
prop_array = np.array([[prop.start_frame, prop.end_frame] for prop in video_pool])
iou_array, overlap_array = segment_tiou(prop_array, prop_array)
self.act_iou_dict[video.id] = torch.from_numpy(iou_array).cuda().float()
# calculate act distance matrix
prop_array = np.array([[prop.start_frame, prop.end_frame] for prop in video_pool])
distance_array = segment_distance(prop_array, prop_array)
self.act_dis_dict[video.id] = torch.from_numpy(distance_array).cuda().float()
pbar.close()
def _prepare_iou_dict(self):
pbar = tqdm(total=len(self.video_list))
for cnt, video in enumerate(self.video_list):
pbar.update(1)
fg = video.get_fg(self.fg_iou_thresh, self.gt_as_fg)
incomp, bg = video.get_negatives(self.incomplete_iou_thresh,
self.bg_iou_thresh,
self.bg_coverage_thresh,
self.incomplete_overlap_thresh)
self.prop_dict[video.id] = [fg, incomp, bg]
video_pool = fg + incomp + bg # 将所有的提议全加到video_pool中
# calculate act iou matrix
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
# 计算所有提议间的IOU和重叠大小(intersection)
iou_array, overlap_array = segment_tiou(prop_array, prop_array)
self.act_iou_dict[video.id] = iou_array
# calculate act distance matrix
# distance = |c1-c2| / Union(c1,c2)
prop_array = np.array([[prop.start_frame, prop.end_frame]
for prop in video_pool])
distance_array = segment_distance(prop_array, prop_array)
self.act_dis_dict[video.id] = distance_array
pbar.close()