Pose Normalization?
jessestauffer opened this issue · comments
Do the source / target individuals need to have similar body sizes / proportions when generating human motion from poses? The subjects appear pretty similar in the demo videos. It is my understanding that extracting poses with OpenPose / DensePose from a short individual would not translate well when generating videos of a large target person. Especially with DensePose since it expresses body mass vs OpenPose just expressing body length.
If so, is there a solution for the pose retargeting built into vid2vid?
Thank you!