3D Gaussian splatting compression method cuts files by 30 times
TSGL: Teacher-Student Graph Learning for 3DGS Compression
Computer Vision and Pattern Recognition
Summary
3D Gaussian Splatting is a way to create 3D images using many small Gaussian shapes, but it produces very large files. The authors propose a new technique called Teacher-Student Graph Learning that compresses these 3D image files without needing to retrain the original model or use the original images. Their method groups similar shapes together in a graph and transforms the data to focus on important details for better compression. They show this method works well on standard tests, shrinking file sizes by about 30 times while keeping the image quality high.
What this means in practice
- •For 3d graphics developers: Shrink large 3D Gaussian splatting models for faster transmission and storage without retraining or needing original images.
- •For video game studios: Reduce file sizes of 3D scene representations to improve game loading times and reduce disk space usage.$Commercial implications: Compression enables smaller game assets with high-quality 3D views, appealing to game studios for distribution and performance benefits.
Authors
Matin Bani Saedi, Matthew Kyan, Gene Cheung
Abstract
3D Gaussian Splatting (3DGS) is a popular representation for novel view synthesis. However, 3DGS contains millions of Gaussian primitives, each with rich attributes, resulting in large file sizes. We propose a novel 3DGS compression method based on Teacher-Student Graph Learning (TSGL) that operates directly on a trained model, without 3DGS retraining or access to training images. Specifically, for each block of Gaussian primitives, using decoded positions and DC spherical harmonic (SH) coefficients as predictors, we learn a signal-dependent geometry graph G encoding the pairwise similarities between neighbouring Gaussians via a teacher-student model. Given G, we perform Graph Fourier Transform (GFT) on the remaining attributes, so that signal energies are predominantly projected into the low-frequency coefficients for compact representation. On three standard benchmarks, the method reaches 27x to 33x compression with less than 0.6 dB of PSNR loss, improving on recent post-training compression methods in both size and rendering quality.