search

Arthur151 / ROMP

Monocular, One-stage, Regression of Multiple 3D People and their 3D positions & trajectories in camera & global coordinates. ROMP[ICCV21], BEV[CVPR22], TRACE[CVPR2023]
https://www.yusun.work/
Apache License 2.0
1.33k stars 229 forks source link

SMPL Model #280

Closed mkhoshle closed 2 years ago

mkhoshle commented 2 years ago

Hi, I have a question regarding smpl model you are using in your code. In the requirements.txt I see that smplx is installed only. However in the paper only SMPL is referenced. Are you using SMPL to create the 3D reconstruction given pose and shape or are you using SMPL-x? If you are using SMPL-x the reconstructed hand gestures should be pretty close to the image, but then why is this not the case looking at the results?

I would appreciate your feedback, Thanks,

Arthur151 commented 2 years ago

Hi, @mkhoshle Good question. Yes, so far, both BEV and ROMP use the SMPL model and estimate the body pose only. The lib smplx contains the implementation of SMPL model, which is employed to use the same evaluation code for a fair comparison with other SOTAs. Best

mkhoshle commented 2 years ago

Thank you very much for the clarification.