Reproduce “Adaptive Patch Exiting for Scalable Single 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
Shizun Wang, Jiaming Liu, Kaixin Chen, Xiaoqi Li, Ming Lu, Yandong Guo
Adaptive Patch Exiting (APE) reduces computation in single-image super-resolution by letting overlapping patches exit a shared network at different depths. A lightweight regressor predicts each patch’s expected improvement from the next layer; an adjustable threshold controls stopping. On DIV2K at ×2, RCAN-APE matches RCAN at 34.09 dB PSNR with 24% of baseline FLOPs; inference-time gains were measured only on an NVIDIA 2080Ti.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.