Reproduce “UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weiqi Li, Xuanyu Zhang, Bin Chen, Jingfen Xie, Yan Wang, Kexin Zhang, Junlin Li, Li Zhang and 2 more
Researchers from Peking University and ByteDance Inc. introduced UARE, a unified vision-language model designed for both image quality assessment (IQA) and image restoration/enhancement (IR/IE) tasks. This model integrates quality analysis with image manipulation, achieving superior perceptual quality and robust performance across diverse degradation types through an assessment-guided restoration approach.
Key insights
- Unifying IQA and IR/IE within a single vision-language model can create a more comprehensive and perceptually aligned image processing system.
- Explicitly leveraging IQA to generate textual guidance during restoration significantly improves the perceptual quality of enhanced images.
- A two-stage training strategy, starting with progressive restoration and then unifying IQA and restoration, effectively balances fidelity and perceptual quality across diverse degradations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.