POTR: Post-Training 3DGS Compression

2026-01-21

AI Summary

Although 3D Gaussian Splatting (3DGS) achieves state-of-the-art performance in novel view synthesis, it suffers from large model size and high storage overhead. To address this, this work proposes a post-processing compression method that requires no retraining. By enhancing the rasterizer to enable efficient parallel pruning and integrating training-free recomputation of lighting coefficients with entropy coding, the approach significantly increases the sparsity of AC coefficients. Without any fine-tuning, the method reduces model size by 2–4×, accelerates rendering by 1.5–2×, and boosts AC coefficient sparsity from 70% to 97%, achieving superior rate-distortion performance compared to existing compression techniques.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

3D Gaussian Splatting (3DGS) has recently emerged as a promising contender to Neural Radiance Fields (NeRF) in 3D scene reconstruction and real-time novel view synthesis. 3DGS outperforms NeRF in training and inference speed but has substantially higher storage requirements. To remedy this downside, we propose POTR, a post-training 3DGS codec built on two novel techniques. First, POTR introduces a novel pruning approach that uses a modified 3DGS rasterizer to efficiently calculate every splat's individual removal effect simultaneously. This technique results in 2-4x fewer splats than other post-training pruning techniques and as a result also significantly accelerates inference with experiments demonstrating 1.5-2x faster inference than other compressed models. Second, we propose a novel method to recompute lighting coefficients, significantly reducing their entropy without using any form of training. Our fast and highly parallel approach especially increases AC lighting coefficient sparsity, with experiments demonstrating increases from 70% to 97%, with minimal loss in quality. Finally, we extend POTR with a simple fine-tuning scheme to further enhance pruning, inference, and rate-distortion performance. Experiments demonstrate that POTR, even without fine-tuning, consistently outperforms all other post-training compression techniques in both rate-distortion performance and inference speed.

Problem

• 3D Gaussian Splatting
• post-training compression
• storage requirements
• real-time rendering
• rate-distortion

Innovation

• 3D Gaussian Splatting
• post-training compression
• pruning
• lighting coefficient sparsity
• rasterization-based analysis