dreamgaussian/dreamgaussian
Dreamgaussian
DreamGaussian is a generative system and converter designed to create textured three-dimensional models from text or images using Gaussian Splatting. It functions as a pipeline for transforming two-dimensional inputs into high-fidelity 3D assets.
The project provides specific workflows for converting 3D Gaussian point clouds into standard textured mesh formats compatible with external 3D software. It supports the generation of textured meshes from single images via volumetric refinement and UV texture optimization, as well as the creation of 3D models from text prompts through intermediate image synthesis.
The capability surface includes mesh quality refinement through iterative training, real-time training visualization, and the rendering of multi-angle preview videos. It also incorporates visual similarity scoring to assess the accuracy of generated objects against reference imagery.
Features
- Gaussian Splatting - Uses Gaussian Splatting to generate high-fidelity 3D assets from images or text.
- Text-to-3D Generators - Synthesizes textured 3D models from natural language descriptions via intermediate image generation.
- 3D Asset Pipelines - Provides a pipeline for processing and converting 3D Gaussian representations into standard formats.
- Image-to-Mesh Generation - Provides the capability to transform a single 2D image into a textured 3D polygonal mesh.
- Gaussian-to-Mesh Converters - Converts 3D Gaussian point clouds into standard textured mesh formats compatible with external software.
- Mesh Format Conversions - Implements a converter that transforms 3D Gaussian point clouds into standard textured mesh formats.
- Text-to-Mesh Generation - Translates natural language descriptions into textured 3D polygonal geometry.
- 3D Asset Previews - Generates 360-degree videos and multi-view image sets to review the quality of generated assets.
- Iterative Geometry Refinement - Refines coarse generative meshes through iterative training to achieve high-fidelity 3D models.
