Fast 2DGS: Efficient Image Representation with Deep Gaussian Prior

2025-12-14

AI Summary

Existing 2D Gaussian Splatting (2DGS) methods rely on post-optimization, with initial Gaussian configurations typically determined by complexity-agnostic heuristics—leading to slow convergence (>10 seconds); learnable initialization networks, in turn, incur additional computational and architectural overhead. Method: We propose a lightweight Deep Gaussian Prior (DGP) framework—the first to decouple conditional generation from dense attribute regression—by modeling Gaussians via spatially aware distribution priors and a compact neural prior, enabling high-fidelity initial Gaussian sets via a single forward pass followed by minimal fine-tuning. Contribution/Results: DGP achieves state-of-the-art visual quality while reducing initialization time to the millisecond level and enabling real-time rendering at over 1000 FPS. This substantially enhances editability and industrial deployability of 2DGS, eliminating the trade-off between quality, speed, and efficiency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

As generative models become increasingly capable of producing high-fidelity visual content, the demand for efficient, interpretable, and editable image representations has grown substantially. Recent advances in 2D Gaussian Splatting (2DGS) have emerged as a promising solution, offering explicit control, high interpretability, and real-time rendering capabilities (>1000 FPS). However, high-quality 2DGS typically requires post-optimization. Existing methods adopt random or heuristics (e.g., gradient maps), which are often insensitive to image complexity and lead to slow convergence (>10s). More recent approaches introduce learnable networks to predict initial Gaussian configurations, but at the cost of increased computational and architectural complexity. To bridge this gap, we present Fast-2DGS, a lightweight framework for efficient Gaussian image representation. Specifically, we introduce Deep Gaussian Prior, implemented as a conditional network to capture the spatial distribution of Gaussian primitives under different complexities. In addition, we propose an attribute regression network to predict dense Gaussian properties. Experiments demonstrate that this disentangled architecture achieves high-quality reconstruction in a single forward pass, followed by minimal fine-tuning. More importantly, our approach significantly reduces computational cost without compromising visual quality, bringing 2DGS closer to industry-ready deployment.

Problem

• Accelerates 2D Gaussian Splatting initialization for efficient image representation
• Reduces computational cost while maintaining high visual quality
• Enables single-pass reconstruction with minimal fine-tuning for industry deployment

Innovation

• Deep Gaussian Prior network predicts Gaussian spatial distributions
• Attribute regression network predicts dense Gaussian properties
• Disentangled architecture enables single-pass reconstruction with minimal fine-tuning