Clean-GS: Semantic Mask-Guided Pruning for 3D Gaussian Splatting

2026-01-01

arXiv.org

AI Summary

This work addresses the challenge of model bloat and target occlusion in 3D Gaussian Splatting (3DGS), which often arises from excessive floating artifacts and cluttered backgrounds, hindering deployment in bandwidth-constrained environments. The authors propose Clean-GS, a novel pruning framework that, for the first time, leverages sparse semantic masks—requiring annotations from only 1% of views—to selectively remove non-target Gaussians. Clean-GS employs a three-stage pipeline: whitelist-based spatial filtering, depth-buffer-guided color consistency verification, and neighborhood-structure anomaly rejection, thereby eliminating reliance on global importance metrics. Evaluated on the Tanks and Temples dataset, the method reduces model size from 125 MB to 47 MB (achieving 60–80% compression) while preserving high-fidelity rendering quality, making it suitable for web and AR/VR applications.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Non-convex Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Data transparency and provenance
• Web Mining and Content Analysis: Web data integration and cleaning

Abstract

3D Gaussian Splatting produces high-quality scene reconstructions but generates hundreds of thousands of spurious Gaussians (floaters) scattered throughout the environment. These artifacts obscure objects of interest and inflate model sizes, hindering deployment in bandwidth-constrained applications. We present Clean-GS, a method for removing background clutter and floaters from 3DGS reconstructions using sparse semantic masks. Our approach combines whitelist-based spatial filtering with color-guided validation and outlier removal to achieve 60-80\% model compression while preserving object quality. Unlike existing 3DGS pruning methods that rely on global importance metrics, Clean-GS uses semantic information from as few as 3 segmentation masks (1\% of views) to identify and remove Gaussians not belonging to the target object. Our multi-stage approach consisting of (1) whitelist filtering via projection to masked regions, (2) depth-buffered color validation, and (3) neighbor-based outlier removal isolates monuments and objects from complex outdoor scenes. Experiments on Tanks and Temples show that Clean-GS reduces file sizes from 125MB to 47MB while maintaining rendering quality, making 3DGS models practical for web deployment and AR/VR applications. Our code is available at https://github.com/smlab-niser/clean-gs

Problem

• 3D Gaussian Splatting
• floaters
• semantic mask
• model compression
• background clutter

Innovation

• Semantic Mask-Guided Pruning
• 3D Gaussian Splatting
• Floaters Removal
• Sparse Semantic Supervision
• Model Compression