Reproduce “Clarity ChatGPT: An Interactive and Adaptive Processing System for Image 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
Yanyan Wei, Zhao Zhang, Jiahuan Ren, Xiaogang Xu, Richang Hong, Yi Yang, Shuicheng Yan, Meng Wang
Researchers introduced Clarity ChatGPT, an interactive and adaptive system integrating a large language model with visual foundation models to address image restoration and enhancement. The system demonstrated enhanced degradation detection (94.57% accuracy), improved image quality through multi-result fusion (22.2% PSNR gain), and enabled multi-round dialogue for iterative, region-specific image processing based on user feedback.
Key insights
- Fine-tuning a CLIP model on a specialized degradation dataset significantly improves its accuracy in classifying diverse image degradation types (from 38.27% to 94.57% Top-1 accuracy).
- Combining outputs from multiple restoration algorithms via a dedicated fusion network can yield superior image quality compared to individual state-of-the-art methods.
- Integrating an LLM for conversational control enables an adaptive, interactive image processing workflow that allows for iterative refinement, region-specific adjustments, and complex chained requests based on user intent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.