Issues with IUV diagrams

[TDIT-haha]

Hello, I have reviewed this paper and think it is done very well.
I have reviewed the issue of generating motion videos using dense poses in iusse and have also attempted to generate them myself. After trying it out on my own, I have a question: can this motion video be generated using a torso segmentation model? If possible, are there any requirements for the color of the mask of the segmented image?