Compression method cuts storage for 3d gaussian splatting models
Towards Practical Compression of 3D Gaussian Splatting
Computer Vision and Pattern Recognition
Summary
3D Gaussian Splatting creates detailed 3D views but the files are very large. The authors introduce COSA-GS, a simpler way to compress these 3D models that avoids complex calculations and errors when decoding on different devices. Their method uses a unique way to look at each 3D point's position for compression and trains the system to keep the quality while reducing data size. This results in faster, reliable, and high-performing compression that works well across platforms.
What this means in practice
- •For 3d graphics developers: Store and deliver high-quality 3D scenes using fewer bytes without risking decoding errors on different user devices.
- •For augmented reality app teams: Reduce data size for detailed real-world scene models so AR experiences load faster and work consistently across devices.
Authors
Pengpeng Yu, Yueru Chen, Fei Song, Tai Qin, Qi Zhang, Jing Wang, Yulan Guo
Abstract
3D Gaussian Splatting (3DGS) enables high-quality novel-view synthesis but requires substantial storage. Existing compression methods often rely on spatial context modeling over irregular 3D representations, increasing the complexity of training and coding. Meanwhile, floating-point context inference can introduce numerical inconsistencies across platforms, causing entropy-decoding failures. To address these practical challenges, we propose COSA-GS, which constructs context without spatial aggregation through anchor-wise causal factorization. Specifically, we use geometry context derived from each anchor's coordinates to model a compact learnable anchor latent. The anchor latent is then fused with the geometry context to form an anchor context for attribute coding. The resulting context model features a simple architecture composed solely of linear transformations and activations. We train COSA-GS using rate--distortion optimization with adaptive Gaussian pruning. Further, we develop quantization-aware training and integer inference for the context model to achieve bit-exact consistency of entropy-decoded symbols across platforms. Experiments demonstrate that COSA-GS achieves state-of-the-art compression performance while retaining fast and consistent cross-platform decoding, providing a simple yet effective framework for practical 3DGS compression. Code is available at https://github.com/pengpeng-yu/COSA-GS.