SkipGS: Post-Densification Backward Skipping for Efficient 3DGS Training

2026-03-09

AI Summary

This work addresses the inefficiency of 3D Gaussian Splatting (3DGS) during its post-densification phase, where training is time-consuming and backpropagation exhibits high redundancy. The authors propose SkipGS, the first method to identify and exploit gradient update redundancy in this stage. SkipGS introduces a plug-and-play, view-adaptive backpropagation gating mechanism that requires no modifications to the renderer or scene representation. By leveraging per-view loss statistics, it selectively skips backpropagation steps yielding minimal improvement, while enforcing a minimum backpropagation budget to ensure optimization stability. Evaluated on the Mip-NeRF 360 dataset, SkipGS reduces end-to-end training time by 23.1% and accelerates the post-densification phase by 42.0%, all while maintaining reconstruction quality comparable to the original 3DGS.

Technology Category

• Computer Vision: 3D Computer Vision
• Search and Optimization: Learning to Search
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

3D Gaussian Splatting (3DGS) achieves real-time novel-view synthesis by optimizing millions of anisotropic Gaussians, yet its training remains expensive, with the backward pass dominating runtime in the post-densification refinement phase. We observe substantial update redundancy in this phase: many sampled views have near-plateaued losses and provide diminishing gradient benefits, but standard training still runs full backpropagation. We propose SkipGS with a novel view-adaptive backward gating mechanism for efficient post-densification training. SkipGS always performs the forward pass to update per-view loss statistics, and selectively skips backward passes when the sampled view's loss is consistent with its recent per-view baseline, while enforcing a minimum backward budget for stable optimization. On Mip-NeRF 360, compared to 3DGS, SkipGS reduces end-to-end training time by 23.1%, driven by a 42.0% reduction in post-densification time, with comparable reconstruction quality. Because it only changes when to backpropagate -- without modifying the renderer, representation, or loss -- SkipGS is plug-and-play and compatible with other complementary efficiency strategies for additive speedups.

Problem

• 3D Gaussian Splatting
• training efficiency
• backward pass
• post-densification
• gradient redundancy

Innovation

• backward skipping
• view-adaptive gating
• 3D Gaussian Splatting
• training efficiency
• post-densification