用变分误差驱动方法实现3D高斯点云的高效压缩，保持实时渲染。
VEDAL: Variational Error-Driven Asynchronous Learning for 3D Gaussian Splatting Pruning
- 基于重建误差异步触发剪枝，动态判断每个高斯点重要性
- 实现5.2倍压缩率，仅损失0.31 dB PSNR，优于现有方法
- 适合需要高保真实时3D渲染的场景，如AR/VR应用
3D高斯泼溅（3DGS）在实现高质量新视角合成与实时渲染方面表现优异，但因数百万个高斯原始体导致内存消耗过大。现有剪枝方法依赖启发式重要性评分或同步批量更新，造成压缩效果不佳且训练不稳定。本文提出VEDAL，将高斯剪枝建模为变分自由能最小化问题。该方法引入(1)预测误差门控机制，根据每个高斯点的重建不确定性异步激活剪枝；(2)变分不确定性头，将剪枝决策视为带有可学习先验的隐变量。自由能目标通过信息论视角自然平衡重建保真度与模型复杂度。在Mip-NeRF 360、Tanks&Temples和Deep Blending数据集上的大量实验表明，VEDAL实现5.2倍压缩，仅损失0.31 dB PSNR，相比PUP 3D-GS在更高压缩比下提升+0.05 dB，相较LightGaussian在相近质量下提升+0.35 dB，同时维持185 FPS的实时渲染性能。
原文摘要 · Abstract (English)
3D Gaussian Splatting (3DGS) achieves remarkable novel view synthesis quality with real-time rendering, yet suffers from excessive memory consumption due to millions of Gaussian primitives. Existing pruning methods rely on heuristic importance scores or synchronous batch updates, leading to suboptimal compression and training instability. We propose VEDAL, a principled framework that formulates Gaussian pruning as variational free energy minimization. Our approach introduces (1) a prediction-error gating mechanism that asynchronously activates pruning based on per-Gaussian reconstruction uncertainty, and (2) a variational uncertainty head that models pruning decisions as latent variables with learnable priors. The free energy objective naturally balances reconstruction fidelity against model complexity through an information-theoretic lens. Extensive experiments on Mip-NeRF 360, Tanks&Temples, and Deep Blending demonstrate that VEDAL achieves 5.2x compression with only 0.31 dB PSNR drop, outperforming PUP 3D-GS by +0.05 dB at a higher compression ratio and LightGaussian by +0.35 dB at comparable quality, while maintaining real-time rendering at 185 FPS.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。