About the DensePose and OpenPose, and how to get the inital "agnostic" mask?

[AndrewChiyz]

Hi, thank you for releasing the code!

I have a few questions about the rendered input of DensePose and OpenPose.

It seems the DensePose and OpenPose as input, is completed and not masked out by using an "agnostic" mask. So I wonder does it mean when testing, only the "agnostic" mask is given to mask out the specific region for virtual try-on? If so, how to get the initial "agnostic" mask?

For DensePose, I wonder how to render and visualize the image of the DensePose IUV map. It seems the I-component stores the semantic labels of 24 human body parts, the U and V components store the U, and V coordinates, respectively. I got some hints from issues #45 and #8, but I am not sure what happens by setting alpha=1. I saw some visualization of the DensePose IUV map. It looks like https://densepose.s3.amazonaws.com/test1uv.0001.png, which is quite different from the images in the folder "image-densepose".

Thanks!