Reproduce “MobiSR: Efficient On-Device Super-Resolution through Heterogeneous Mobile Processors” 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
Royson Lee, Stylianos I. Venieris, Łukasz Dudziak, Sourav Bhattacharya, Nicholas D. Lane
MobiSR routes patches between differently sized super-resolution models across compute engines, using total variation as a proxy for upscaling difficulty. It sends easier patches to the larger model and harder patches to the compact model. On B100, it was up to 94% faster than a selected single-model baseline at the same average PSNR-drop budget. Tests covered one Snapdragon 845 platform.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.