search

mks0601 / 3DMPPE_POSENET_RELEASE

Official PyTorch implementation of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image", ICCV 2019
MIT License
806 stars 147 forks source link

about h36m parsing. #126

Open llpspark opened 1 year ago

llpspark commented 1 year ago

Hi ,wonderful work for 3D human pose estimation. I wonder how to parsing the raw h36m data from pose to swing and twist. thank you.

mks0601 commented 1 year ago

In this repo, I provided only 3D joint coordinates, not swing and twist rotations. For the SMPL pose parameters (https://drive.google.com/drive/folders/1ySxiuTCSdUEqbgTcx7bx02uMglPOkKjc?usp=sharing), which include rotations, I used NeuralAnnot (my previous work) to get 3D pseudo-GTs. Please refer to that paper.