Reproduce “TSGL: Teacher-Student Graph Learning for 3DGS Compression” 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
Matin Bani Saedi, Matthew Kyan, Gene Cheung
TSGL compresses already-trained 3DGS scenes without retraining or original training images. It builds a graph from attribute similarities, then applies a graph Fourier transform and entropy coding to Gaussian attributes. Across 13 scenes in three benchmarks, models are 27×–33× smaller with average PSNR loss below 0.6 dB. PSNR measures rendered-image fidelity; results apply to these tested datasets and pipeline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.