Reproduce “TaskIR: Task-Driven Image Restoration via Degradation Adaptation and Task Feedback” 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
Yanjie Tu, Qingsen Yan, Axi Niu, Wenxuan Cai, Tao Hu, Wei Dong, Haokui Zhang
TaskIR addresses image restoration across eight degradation types while preserving information for downstream vision tasks. It first uses global and spatial degradation representations to adapt repairs, then applies task feedback selectively. On Cityscapes, it reaches 35.31 dB PSNR; full feedback raises mIoU from 74.41 to 74.69 versus no-feedback ablation, with comparable restoration quality. It does not lead on VOC2012 [email protected]:0.95.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.