MarkSplatter: Generalizable Watermarking for 3D Gaussian Splatting Model via Splatter Image Structure

2025-08-31

AI Summary

To address the lack of efficient, general-purpose watermarking schemes for 3D Gaussian Splatting (3DGS) models, this paper proposes the first fine-tuning-free, single-forward watermarking framework enabling both embedding and extraction. The method comprises two core innovations: (1) GaussianBridge, a novel mapping that transforms unstructured 3D Gaussian point clouds into splatter images amenable to neural processing; and (2) a Gaussian uncertainty-aware heatmap prediction module coupled with dense segmentation, enabling robust watermark extraction from small localized regions. The framework preserves rendering quality while allowing rapid, message-agnostic watermark embedding. Experimental results show near-perfect extraction accuracy (>99.5%) even under extremely low spatial coverage (<1%), significantly outperforming existing approaches that require per-message fine-tuning. This marks the first generalizable, lightweight watermarking solution for 3DGS, advancing copyright protection for emerging 3D generative models.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Data transparency and provenance

Abstract

The growing popularity of 3D Gaussian Splatting (3DGS) has intensified the need for effective copyright protection. Current 3DGS watermarking methods rely on computationally expensive fine-tuning procedures for each predefined message. We propose the first generalizable watermarking framework that enables efficient protection of Splatter Image-based 3DGS models through a single forward pass. We introduce GaussianBridge that transforms unstructured 3D Gaussians into Splatter Image format, enabling direct neural processing for arbitrary message embedding. To ensure imperceptibility, we design a Gaussian-Uncertainty-Perceptual heatmap prediction strategy for preserving visual quality. For robust message recovery, we develop a dense segmentation-based extraction mechanism that maintains reliable extraction even when watermarked objects occupy minimal regions in rendered views. Project page: https://kevinhuangxf.github.io/marksplatter.

Problem

• Protecting copyright of 3D Gaussian Splatting models efficiently
• Embedding watermarks without expensive per-message fine-tuning
• Ensuring imperceptible yet robust watermark extraction

Innovation

• Generalizable watermarking via single forward pass
• GaussianBridge transforms Gaussians to Splatter Image
• Dense segmentation-based extraction for robust recovery