search

mks0601 / Hand4Whole_RELEASE

Official PyTorch implementation of "Accurate 3D Hand Pose Estimation for Whole-Body 3D Human Mesh Estimation", CVPRW 2022 (Oral.)
MIT License
315 stars 31 forks source link

video demo #42

Open dtgfdgfgf opened 2 years ago

dtgfdgfgf commented 2 years ago

Hi, I am trying to use a video as an input in demo.py. I hope finally I can get a video that a model make same action as a human. Therefore, I have to get bbox frame by frame, but bbox is manually set in demo.py line60. I am wondering if I can call the function get_bbox() in ./utils/preprocessing.py to get bbox. If get_bbox() actually works, then how to set the arguments "joint_img、joint_valid"?

Or is there any suggestion for using a video to demo?

mks0601 commented 2 years ago

Use human detection or tracking networks like Yolo v5

dtgfdgfgf commented 2 years ago

thank you!