SymGS : Leveraging Local Symmetries for 3D Gaussian Splatting Compression

TL;DR

SymGS tackles the memory bottleneck of 3DGS by exploiting reflective symmetries in scenes. It introduces a CUDA-accelerated symmetry detector and a hierarchical mirror-based compression framework, enabling large storage reductions while preserving rendering quality. The method is designed as a plug-in to the HAC compression pipeline and yields significant gains across diverse datasets, especially in symmetry-rich environments. The approach preserves rendering speed due to pre-computed symmetry handling and provides a more interpretable scene decomposition.

Abstract

3D Gaussian Splatting has emerged as a transformative technique in novel view synthesis, primarily due to its high rendering speed and photorealistic fidelity. However, its memory footprint scales rapidly with scene complexity, often reaching several gigabytes. Existing methods address this issue by introducing compression strategies that exploit primitive-level redundancy through similarity detection and quantization. We aim to surpass the compression limits of such methods by incorporating symmetry-aware techniques, specifically targeting mirror symmetries to eliminate redundant primitives. We propose a novel compression framework, SymGS, introducing learnable mirrors into the scene, thereby eliminating local and global reflective redundancies for compression. Our framework functions as a plug-and-play enhancement to state-of-the-art compression methods, (e.g. HAC) to achieve further compression. Compared to HAC, we achieve $1.66 \times$ compression across benchmark datasets (upto $3\times$ on large-scale scenes). On an average, SymGS enables $\bf{108\times}$ compression of a 3DGS scene, while preserving rendering quality. The project page and supplementary can be found at symgs.github.io

Figures (9)

• Figure 1: Our SymGS leverages Reflective Symmetries in a 3DGS scene for compression while preserving rendering quality.
• Figure 2: PSNR vs RCF for 3DGS compression.
• Figure 3: SymGS Framework: Given a 3DGS scene, we first perform Gaussians clustering and voting to identify the dominant mirror symmetry. Then, the Gaussians on one side of the mirror are replaced with reflections of their counterparts. Finally, the modified Gaussian set and mirror parameters are jointly optimized using the standard photometric loss.
• Figure 4: Iterative Symmetry Detection & Optimization.
• Figure 5: Mirror-aware optimization restores the visual details which might get attenuated while reflecting the Gaussians.