CG3D: Compositional Generation for Text-to-3D via Gaussian Splatting

This paper introduces CG3D, a novel method for generating scalable 3D assets from text prompts. Current text-to-3D methods struggle to produce detailed, multi-object scenes and maintain control over object configurations. CG3D addresses these issues by using explicit Gaussian radiance fields, allowing for the creation of semantically and physically consistent scenes. The method also boasts superior performance in terms of object combinations and physics accuracy compared to the guiding diffusion model.

Publication date: 29 Nov 2023
Project Page: https://asvilesov.github.io/CG3D/
Paper: https://arxiv.org/pdf/2311.17907