Authors: Yuanxing Duan, Fangyin Wei, Qiyu Dai, Yuhang He, Wenzheng Chen, Baoquan Chen

Published on: February 05, 2024

Impact Score: 8.07

Arxiv code: Arxiv:2402.03307

Summary

• What is new: Introduction of 4D Gaussian Splatting (4DGS) for novel view synthesis in dynamic scenes, providing a novel approach that excels in capturing high-fidelity renderings and handling abrupt motions.
• Why this is important: Current methods for novel view synthesis in dynamic scenes struggle with sudden movements and achieving high-fidelity renderings.
• What the research proposes: 4DGS represents dynamic scenes with anisotropic 4D XYZT Gaussians, using temporal slicing to model dynamics and enable seamless image projection.
• Results: Achieved real-time inference rendering speeds of up to 277 FPS on an RTX 3090 GPU and 583 FPS on an RTX 4090 GPU, with superior efficiency and effectiveness demonstrated across various scenes.

Technical Details

Technological frameworks used: CUDA acceleration

Models used: 4D Gaussian Splatting (4DGS)

Data used: Diverse motion scenes

Potential Impact

VFX and game development industries, AR/VR technology companies

Want to implement this idea in a business?

We have generated a startup concept here: GaussScene Dynamics.