Skip to content

Latest commit

 

History

History
24 lines (12 loc) · 863 Bytes

README.md

File metadata and controls

24 lines (12 loc) · 863 Bytes

MeshLifter

MeshLifter: Weakly supervised approach for 3D human mesh reconstruction from a single 2D pose based on loop structure

To download data files like annotation for Human3.6M https://drive.google.com/drive/folders/1aWzMfDqTOAVnSxYB2Gc_XoyfgW6m653v?usp=sharing

And Pretrained Model https://drive.google.com/file/d/1ZZZJ55vJ8fdRmgzepnHHZYAm_Xm7eRj3/view?usp=sharing

Human3.6M dataset and Parsed file https://drive.google.com/drive/folders/1kgVH-GugrLoc9XyvP6nRoaFpw3TmM5xK

demo.py : To visualize 2D pose(input), reconstructed 3D Mesh and corresponding 2D images(Not input)

test.py : To quantitatively evaluated this project

And you have to change the path of annoatation file, pretrained model and smpl files..