PointSplat: Efficient Geometry-Driven Pruning and Transformer Refinement for 3D Gaussian Splatting

2026-04-10

AI Summary

This work addresses the high memory and storage costs of 3D Gaussian Splatting (3DGS), which relies on millions of Gaussians to represent complex scenes. The authors propose an efficient compression framework that operates without 2D image guidance or per-scene fine-tuning. Their approach first prunes redundant Gaussians based solely on 3D geometric attributes, substantially reducing representation size. Subsequently, a dual-branch Transformer architecture is introduced to separately refine geometric and appearance features, mitigating feature imbalance between the two modalities. Evaluated on the ScanNet++ and Replica datasets, the method achieves rendering quality comparable to state-of-the-art approaches across various sparsity levels while significantly improving computational and storage efficiency.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Learning on the Edge & Model Compression
• Data Mining & Knowledge Management: Data Compression

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 unlocked real-time, high-fidelity novel view synthesis by representing scenes using explicit 3D primitives. However, traditional methods often require millions of Gaussians to capture complex scenes, leading to significant memory and storage demands. Recent approaches have addressed this issue through pruning and per-scene fine-tuning of Gaussian parameters, thereby reducing the model size while maintaining visual quality. These strategies typically rely on 2D images to compute important scores followed by scene-specific optimization. In this work, we introduce PointSplat, 3D geometry-driven prune-and-refine framework that bridges previously disjoint directions of gaussian pruning and transformer refinement. Our method includes two key components: (1) an efficient geometry-driven strategy that ranks Gaussians based solely on their 3D attributes, removing reliance on 2D images during pruning stage, and (2) a dual-branch encoder that separates, re-weights geometric and appearance to avoid feature imbalance. Extensive experiments on ScanNet++ and Replica across varying sparsity levels demonstrate that PointSplat consistently achieves competitive rendering quality and superior efficiency without additional per-scene optimization.

Problem

• 3D Gaussian Splatting
• model compression
• memory efficiency
• scene representation
• pruning

Innovation

• 3D Gaussian Splatting
• geometry-driven pruning
• Transformer refinement
• dual-branch encoder
• novel view synthesis