Question about optimized body pose

tijiang13 / InstantAvatar

InstantAvatar: Learning Avatars from Monocular Video in 60 Seconds (CVPR 2023)

365 stars 31 forks source link

Question about optimized body pose #65

Closed TohnoYouki closed 5 months ago

TohnoYouki commented 8 months ago

Thanks for your excellent work. I found that there is a significant difference in the results without provided optimized poses in peoplesnapshot dataset. I tried to optimize SMPL parameters in advance with fit.py following the procedure in run-neuman-demo.sh but it didn't make any difference. So how can I train to get the optimized poses which are same to your provided ? Thanks.

tijiang13 commented 7 months ago

Hi TohnoYouki,

We obtained the 'optimized poses' by refining the poses provided by PeopleSnapshot, which may contain noise in foot estimation. However, if you have already executed the fit.py script on the raw data beforehand, the results should not vary significantly. It would be beneficial if you could provide screenshots along with the scripts you are using.

Best, Tianjian