Reproduce “DNF-SR: Dual-Input and Negative-Aware Feature Fine-Tuning for Real-World Image Super-Resolution” 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
Shuhao Han, Wenjie Liao, Hayden Vance, Hang Dong, Rui Zhang, Chun-Le Guo, Chongyi Li
DNF-SR restores degraded low-resolution photos in one diffusion step by feeding an image-editing model both a noisy low-resolution latent and the untouched input, then refining outputs with reward-based, image-space post-training. On RealSR, it reaches a CLIPIQA score of 0.7257 versus 0.7196 for TSDSR. This no-reference gain does not establish universal fidelity: reference-based results are mixed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.