search

sangyun884 / HR-VITON

Official PyTorch implementation for the paper High-Resolution Virtual Try-On with Misalignment and Occlusion-Handled Conditions (ECCV 2022).
825 stars 171 forks source link

Bad results on some images. #46

Open OliverVarnce opened 1 year ago

OliverVarnce commented 1 year ago

Who can explain to me why in some images I get such a result. The image of agnostics is superimposed on the output image.

00095_00_00095_00

What are the requirements for models images?

ds-nat-j commented 1 year ago

Even I'm facing the same issue, I've used just changed the cloth from a custom dataset and applied cloth-segmentation to get the cloth-mask. I've used the same human model with as they have but the results look disastrous.

But if I use a cloth from their test dataset, it look clean. Leads me to believe that they've used their test dataset in training and this paper is rigged. Please prove me wrong.

Screenshot 2023-04-03 at 2 11 53 PM
AI1120 commented 11 months ago

I faced the same issue. How can I solve it?

curious-ai-developer commented 5 months ago

Plea tell me how I can get results like you got with a custom cloth on their human model.