Reproduce “GigaBench: Benchmarking Learnable Representations for Gigapixel Image Compression and 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

Matteo Fusconi, Alex Costanzino, Pierluigi Zama Ramirez, Luigi Di Stefano

GigaBench compares existing learnable representations for compression and super-resolution on ten images exceeding one gigapixel each. It adapts methods for scale and measures quality, compression, and runtime. At similar 5.1× compression, MetricGrids averages 33.59 dB PSNR versus I-NGP’s 33.19 dB, but takes longer; human rankings favor classical interpolation at ×4, with hybrids close at ×8. These findings apply only to tested methods and scenes.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.