3D Gaussian Splatting

About 3D Gaussian Splatting

3D Gaussian Splatting is a groundbreaking tool developed by Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. It was presented at SIGGRAPH 2023 and published in ACM Transactions on Graphics. This tool has revolutionized the field of novel-view synthesis of scenes captured with multiple photos or videos.

Background and Development

The tool was developed by a team of researchers from Inria, Université Côte d'Azur, and MPI Informatik. It represents a significant milestone in the field of radiance field methods, offering state-of-the-art visual quality while maintaining competitive training times.

Core Features and Capabilities

3D Gaussian Splatting introduces three key elements that allow it to achieve high-quality real-time novel-view synthesis at 1080p resolution. These include the representation of the scene with 3D Gaussians, interleaved optimization/density control of the 3D Gaussians, and a fast visibility-aware rendering algorithm.

User Experience

The tool offers a user-friendly interface and a seamless user experience. It supports anisotropic splatting and accelerates both training and real-time rendering.

Applications and Use Cases

3D Gaussian Splatting can be applied across different scenarios or sectors, including professional settings, personal use, and educational purposes. It has been tested on a total of 13 real scenes taken from previously published datasets and the synthetic Blender dataset.

Impact and Future Outlook

3D Gaussian Splatting has had a significant impact on its users and the industry. It has set a new standard in NeRF rendering quality and continues to evolve based on user feedback and technological advancements.