它的跟踪技术
首先在当前帧选一个人体,跟前一帧所有人比较,如果相似度大于阈值,则把前一帧相似度最大人的序号赋予当前帧,且之后当前帧其他人不与其进行相似度估计。
current_poses = sorted(current_poses, key=lambda pose: pose.confidence, reverse=True) # match confident poses first
mask = np.ones(len(previous_poses), dtype=np.int32)
flag=0
for current_pose in current_poses:
best_matched_id = None
best_matched_pose_id = None
best_matched_iou = 0
all_image=0
for id, previous_pose in enumerate(previous_poses):
if not mask[id]:
continue
iou = get_similarity(current_pose, previous_pose)
if iou > best_matched_iou:
best_matched_iou = iou
best_matched_pose_id = previous_pose.id
best_matched_id = id
all_image=previous_pose.all_save_image
if best_matched_iou >= threshold:
mask[best_matched_id] = 0
else: # pose not similar to any previous
best_matched_pose_id = None
all_image=None
如果要去解决遮挡问题的话,直接用一个pose跟之前所有的进行相似度计算因此不用mask。
- lightweight openposepose_entries lightweight openpose entries lightweight openpose openpose openpose论文 lightweight openpose inpaint control net self-supervised transformers lightweight supervised applications lightweight continuous framework lightweight translating pseudocode mechanism openpose人体 系统