Results are larger but not improved?

XPixelGroup / HAT

CVPR2023 - Activating More Pixels in Image Super-Resolution Transformer Arxiv - HAT: Hybrid Attention Transformer for Image Restoration

Apache License 2.0

1.18k stars 140 forks source link

Results are larger but not improved? #2

Open AIisCool opened 2 years ago

AIisCool commented 2 years ago

Input:

Output:

chxy95 commented 2 years ago

@AIisCool The provided pretrained models aim to perform High-Quality Image Super-Resolution, for which can not deal with other degradations like compression or blurring. We would consider training a HAT model that can handle multiple degradations, but no model is available yet currently. This project might help Real-ESRGAN.