This Demo is based on yolov5s for hand detection, and lightweight-human-pose-estimation for hand 21 keypoints estimation yolov5 github:https://github.com/ultralytics/yolov5 lightweight-human-pose-estimation:https://github.com/Daniil-Osokin/lightweight-human-pose-estimation.pytorch

At the end of the hand keypoints network, a few layers of network are added to directly return to 21 hand keypoints

How to run the algorithm

Note:The weight of checkpoint_iter_1100000.pth does not include the structure of the direct return to 21 hand keypoints, only the heatmap can be calculated

run demo --python detect.py

if you want to use heatmap to calculate hand keypoints,--use_heatmap==True,--hand_pose_weight use checkpoint_iter_105000.pth or checkpoint_iter_1100000.pth

if you want to directly return to 21 hand keypoints,--use_heatmap==True,--hand_pose_weight only use checkpoint_iter_105000.pth

bilibili video : https://space.bilibili.com/515212620?spm_id_from=333.788.b_765f7570696e666f.2