Dense_pose setting for creating dataset

[arhamlet]

I have tried to run both the detectron and detectron2 repositories. I got densepose IUV and INDS images for the VITON images and compared them with the .npy files available in the dataset. However, they don't match. Can you kindly tell me what settings of densepose do you run and also how you convert the output image to numpy. I know the numpy conversion but what channels or shape do you choose because the simple numpy comversion with uint8 gives a numpy file that does not match with the original one.

[wang674]

Hello, I also encountered this problem. Have you solved it？ @arhamlet