Recent advances in 3D Gaussian splatting

Analysis

Why This Paper Matters

This survey arrives at a critical juncture in the rapidly evolving field of 3D scene representation. 3D Gaussian splatting (3DGS) has emerged as a powerful alternative to neural radiance fields (NeRFs), offering dramatically faster rendering while maintaining high quality. The paper's comprehensive categorization of methods by functionality—reconstruction, editing, and downstream applications—provides a much-needed roadmap for both newcomers and seasoned researchers. By systematically organizing the growing body of work, it helps the community understand the landscape, identify gaps, and build upon existing approaches.

The significance is amplified by the explicit nature of 3DGS representation, which enables not only fast rendering but also facilitates downstream tasks like dynamic reconstruction, geometry editing, and physical simulation. This survey highlights how 3DGS is moving beyond pure novel view synthesis into interactive and creative applications, signaling a shift toward more practical and deployable 3D technologies.

Technical Contributions

The paper's main technical contribution is its structured taxonomy of 3DGS methods:

• 3D Reconstruction: Methods that use 3DGS for scene reconstruction from images or video, including static and dynamic scenes.
• 3D Editing: Techniques that leverage the explicit Gaussian representation for geometry editing, appearance editing, and composition.
• Downstream Applications: Uses in physical simulation, robotics, and other areas that benefit from the explicit, differentiable representation.
• Foundations: A review of traditional point-based rendering and the rendering formulation of 3DGS, providing essential background.

Results

As a survey, the paper does not present new experimental results. However, it synthesizes key findings from the literature, noting that 3DGS achieves real-time rendering at high visual fidelity, often surpassing NeRFs in speed by orders of magnitude while maintaining comparable or better quality. The survey references 299 citations, indicating the breadth and rapid growth of the field.

Significance

This survey serves as a definitive reference for the 3DGS community, consolidating a fragmented literature into a coherent framework. It will likely accelerate research by helping practitioners quickly understand the state of the art and identify promising directions. The explicit representation of 3DGS opens doors to interactive applications in VR/AR, gaming, and robotics, where real-time performance is critical. By highlighting the versatility of 3DGS beyond rendering, the paper encourages exploration of new use cases, potentially expanding the impact of neural rendering techniques into broader AI and graphics domains.