3DGS Compression
3DGS: 3D Gaussian Splatting
Overview
3D Gaussian Splatting compression studies how to represent scenes with fewer or cheaper-to-store Gaussian primitives without losing important visual detail. Common approaches prune or merge redundant splats, allocate capacity by importance, quantize attributes, and use anchors, entropy models, or image and video codecs to encode what remains. Some methods compress dynamic scenes by sharing stable content across frames or coding motion and residuals, with evaluation spanning novel-view benchmarks such as Mip-NeRF 360 and streaming or mobile settings. The central tension is balancing rate and rendering quality across viewpoints and motion while keeping encoding, decoding, and quality assessment practical.
Latest papers
85 papers17 papers in the last four weeks, up 13% on the four weeks before. 0.1% of all new papers.