2023/11/21 by Kai Katsumata, Katsumata, Kai, Duc Minh Vo +3 · 22 citations
Computer Science · #Advanced Image Processing Techniques #Advanced Vision and Imaging #Artificial intelligence #Computer Graphics and Visualization Techniques #Computer graphics (images) #Computer science #Computer vision #Gaussian #Monocular #Real-time rendering #Rendering (computer graphics) #Representation (politics)
paper · pdf · doi:10.48550/arxiv.2311.12897
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2023/11/21 · openalex created_date 2023/11/24 · openalex updated_date 2026/08/05
3D Gaussian Splatting (3DGS) has shown remarkable success in synthesizing novel views given multiple views of a static scene. Yet, 3DGS faces challenges when applied to dynamic scenes because 3D Gaussian parameters need to be updated per timestep, requiring a large amount of memory and at least a dozen observations per timestep. To address these limitations, we present a compact dynamic 3D Gaussian representation that models positions and rotations as functions of time with a few parameter approximations while keeping other properties of 3DGS including scale, color and opacity invariant. Our method can dramatically reduce memory usage and relax a strict multi-view assumption. In our experiments on monocular and multi-view scenarios, we show that our method not only matches state-of-the-art methods, often linked with slower rendering speeds, in terms of high rendering quality but also significantly surpasses them by achieving a rendering speed of 118 frames per second (FPS) at a resolution of 1,352×1,014 on a single GPU.