image agnostic wrong!!!

[appleyang123]

When i use your data_loader.py to generate the image agnostic in img_agnostic function. The result is wrong?
The first image is the model image. The second image is the cloth image. And the third image is the image agnostic returned by img_agnostic function.

[appleyang123]

[lyc0929]

When i use your data_loader.py to generate the image agnostic in img_agnostic function. The result is wrong? The first image is the model image. The second image is the cloth image. And the third image is the image agnostic returned by img_agnostic function.

please check your function's inputs ,the function can process the mask correctly when I use VITON-HD dataset.

[appleyang123]

I have not modified your dataloader.py. In your original getitem function, it calls the get_img_agnostic function.
You can also check it.

[appleyang123]

Maybe the VITON-HD used in my version is not consistent with yours. It seems that something (VITON-HD train/image-parse-v3 and train/openpose_json) are not consistent.

[rohitpaul23]

@appleyang123 Have you solved the problem?