3D Gaussian Splatting: Survey, Technologies, Challenges, and Opportunities

Yanqi Bao,Tianyu Ding,Jing Huo,Yaoli Liu,Yuxin Li,Wenbin Li,Yang Gao,Jiebo Luo

2024-07-25

Abstract:3D Gaussian Splatting (3DGS) has emerged as a prominent technique with the potential to become a mainstream method for 3D representations. It can effectively transform multi-view images into explicit 3D Gaussian representations through efficient training, and achieve real-time rendering of novel views. This survey aims to analyze existing 3DGS-related works from multiple intersecting perspectives, including related tasks, technologies, challenges, and opportunities. The primary objective is to provide newcomers with a rapid understanding of the field and to assist researchers in methodically organizing existing technologies and challenges. Specifically, we delve into the optimization, application, and extension of 3DGS, categorizing them based on their focuses or motivations. Additionally, we summarize and classify nine types of technical modules and corresponding improvements identified in existing works. Based on these analyses, we further examine the common challenges and technologies across various tasks, proposing potential research opportunities.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

The paper primarily focuses on the development of **3D Gaussian Splatting (3DGS)** technology and its applications in various fields. Specifically, the paper attempts to:

1. **Systematically summarize and categorize work related to 3DGS**:
- Provide a comprehensive review covering the optimization, application, and extension of 3DGS.
- Offer a detailed classification and improvement of existing technical modules, such as initialization, attribute setting, regularization, etc.

2. **Analyze the commonalities and challenges of 3DGS in different tasks**:
- Explore the applications of 3DGS in virtual reality (VR), augmented reality (AR), robotics, movie special effects, and other fields.
- Systematically summarize the common challenges of 3DGS in different tasks and propose potential research opportunities.

3. **Propose core challenges and future research directions for 3DGS**:
- Identify issues in storage efficiency, training efficiency, rendering efficiency, and propose corresponding solutions.
- Highlight the limitations of existing research and propose new ways to address core challenges and advance the field.

4. **Promote the popularization and development of 3DGS technology**:
- Release an open-source project that compiles articles and technologies related to 3DGS, enabling more researchers to access the latest research findings.

Through these efforts, the paper aims to help readers quickly understand the current state of 3DGS development and provide valuable references for future research.