#3D Gaussians

Discover S3Gaussian's approach using 3D Gaussians for self-supervised street scene modeling in autonomous driving, bypassing traditional 3D bounding boxes. It features an innovative hexplane-based encoder and a multi-head Gaussian decoder for quality scene rendering. Compatible with Ubuntu, Python, and PyTorch, this open-source initiative offers extensive tools for training and visualization, highlighting its advancements in modeling dynamic environments without extra supervision.

Explore the Gaussian Opacity Fields (GOF) project, which utilizes 3D Gaussians and Marching Tetrahedra to streamline surface reconstruction in expansive environments. This technique enhances mesh extraction efficiency and speed, with recent advancements doubling training pace through optimized operations. Featured at SIGGRAPH ASIA 2024, GOF demonstrates exceptional performance on datasets such as TNT and Mip-NeRF 360.

Explore a state-of-the-art simulation-rendering framework that effectively utilizes video diffusion models for decoding the dynamics of 3D Gaussian physics. The project includes available training code, synthetic datasets, and forthcoming additions such as tutorials and demonstrations. Benefit from pre-trained 3D Gaussian model integration and text-to-video diffusion model optimization for enhanced physical parameter tuning. This tool, underpinned by valuable open-source methodologies and datasets, serves as a robust platform for advancing physics simulation research.

GaussianShader optimizes neural rendering for reflective surfaces using a simplified shading function applied to 3D Gaussians. This approach ensures high material and lighting fidelity at real-time speeds. Starting with neural 3D Gaussian spheres, it integrates shading attributes for precise view-dependent appearance capture. It also includes a differentiable environment lighting map to facilitate free-viewpoint rendering in diverse lighting conditions, proving effective on datasets like NeRF Synthetic and Tanks&Temples.