3D Gaussian splatting method cuts data size over 150 times
RacerGS: Residual-Autoregressive Context for 3D Gaussian Splatting Compression
Computer Vision and Pattern RecognitionGraphics
Summary
3D Gaussian Splatting is a way to create 3D images quickly and clearly, but it needs a lot of storage space. The authors introduce RacerGS, a new method that compresses these 3D images to much smaller sizes without losing image quality. It uses smart features like hash projections and a special model that predicts and compresses data efficiently by understanding patterns in the 3D scene. Tests showed RacerGS can shrink data size by over 150 times compared to the original method while keeping images sharp and detailed.
What this means in practice
- •For 3d graphics engineers: Store and transmit large 3D scenes using RacerGS to save storage and bandwidth while maintaining high-quality rendering for real-time applications.
- •For virtual reality developers: Optimize 3D environment data size in VR systems to reduce loading times and improve streaming performance without degrading visual fidelity.
Authors
Wenqing Wang, Huimin Zeng, Yun Fu
Abstract
3D Gaussian Splatting (3DGS) has emerged as a promising approach for fast, high-quality novel-view rendering. However, its large parameter size remains a bottleneck for storage and transmission. We propose RacerGS, a compact 3DGS compression framework that achieves scene reconstruction at significantly reduced storage sizes with preserved fidelity. To obtain compact and spatially coherent features, we design a hash projection model that projects the anchor hash encodings into compact context features. Additionally, we introduce a FiLM-modulated autoregressive GRU context model that utilizes causal dependencies among anchor feature groups and uses recurrent hidden states to refine entropy-model parameters, improving bitrate efficiency. Furthermore, we adaptively entropy-code quantized anchor-attribute residuals relative to their context-predicted means, leveraging spatial consistency to reduce symbol entropy. Overall, RacerGS achieves more than 150x compression over vanilla 3DGS and 28x over Scaffold-GS, while providing comparable rendering quality. Extensive experiments across five datasets demonstrate that RacerGS outperforms prior 3DGS compression methods in both storage reduction and rate-distortion performance.