search

Jeff-sjtu / HybrIK

Official code of "HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation", CVPR 2021
MIT License
1.16k stars 142 forks source link

Change number of joint keypoint from 29 to 17 #167

Open pinya52 opened 1 year ago

pinya52 commented 1 year ago

Hi, thanks for your work.

I have successfully replicated your work, and it is truly impressive. However, I have encountered an issue regarding multiperson videos. The pose estimation becomes unstable when there are multiple people in the video. To address this, I have attempted to track the skeleton of each person using Alphapose and MotionBert, which resulted in obtaining 17 3D joint keypoints. I am curious to know if this work can still produce the same result even if the input number of joint keypoints is not 29. Thanks.

AndreOliveira00 commented 1 year ago

Hi @pinya52,

As the author said in this issue #121, the model also gives the format of 24/17 joints output.