RAVE: Rate-Adaptive Visual Encoding for 3D Gaussian Splatting

2025-12-07

AI Summary

To address the high memory footprint and training cost of 3D Gaussian Splatting (3DGS), as well as the limitation of existing compression methods—supporting only fixed bitrates and thus lacking adaptability to heterogeneous devices and variable-bandwidth scenarios—this paper proposes the first lightweight, rate-adaptive compression framework for 3DGS. Our method jointly optimizes rate-distortion performance via differentiable quantization and introduces a compact encoding architecture, enabling continuous bitrate interpolation over 3D Gaussian point clouds for the first time: arbitrary bitrates within a predefined range can be achieved without retraining, while preserving high-fidelity rendering quality. Experiments demonstrate that our approach significantly outperforms fixed-bitrate baselines across a wide bitrate spectrum, achieving both high compression efficiency and real-time inference capability. This work provides a flexible, efficient solution for cross-platform deployment of immersive multimedia applications.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Recent advances in neural scene representations have transformed immersive multimedia, with 3D Gaussian Splatting (3DGS) enabling real-time photorealistic rendering. Despite its efficiency, 3DGS suffers from large memory requirements and costly training procedures, motivating efforts toward compression. Existing approaches, however, operate at fixed rates, limiting adaptability to varying bandwidth and device constraints. In this work, we propose a flexible compression scheme for 3DGS that supports interpolation at any rate between predefined bounds. Our method is computationally lightweight, requires no retraining for any rate, and preserves rendering quality across a broad range of operating points. Experiments demonstrate that the approach achieves efficient, high-quality compression while offering dynamic rate control, making it suitable for practical deployment in immersive applications. The code will be provided open-source upon acceptance of the work.

Problem

• Compresses 3D Gaussian Splatting to reduce memory and training costs
• Enables flexible rate adaptation for varying bandwidth and device constraints
• Maintains rendering quality without retraining across different compression rates

Innovation

• Flexible compression scheme for 3D Gaussian Splatting
• Supports interpolation at any rate without retraining
• Lightweight method preserving rendering quality across rates