Closed tyugv closed 1 year ago
Hi! Sorry for bothering, I figured it out.
Earlier I resized all the images to small size before running the code but it is better to save them with resolution 750x1011 and then use transforms.Resize. For keypoints I changed orig size in load_pose_from_json like here https://github.com/cuiaiyu/dressing-in-order/issues/36#issuecomment-1047393508 . The result was a bit better this way.
Also, I noticed that I had extra numbers in order when I called dress_in_order. It had a bad impact on the results.
Here are my final results:
Hi! Thank you very much for your work!
I tried some garments from the demo notebook with an image of another person. These are my results:
![image](https://github.com/cuiaiyu/dressing-in-order/assets/55535127/0cc42ab3-add7-45a4-b318-d4a5e56e4425)
I noticed that clothing items look blurry compared to the ones from the notebook. Is it an expected result?![image](https://github.com/cuiaiyu/dressing-in-order/assets/55535127/42328629-f6a9-4703-93f5-7aa68e04ac8b)
Maybe the problem is in my code. I changed it to use try on with custom images