kaolin

Why we included this project

Kaolin, NVIDIA's PyTorch library for 3D deep learning, packs a lot of research-grade pieces into one package: GPU-optimized modules for differentiable rendering, physics simulation, and 3D Gaussian splat handling, plus a GPU octree acceleration structure for point clouds. You also get fast conversions between mesh, voxel, and point-cloud forms, along with quaternion math and PLY/USD I/O, so you don't have to write those kernels yourself. The tutorials and documentation lower the barrier for reproducing published work or trying out new ideas. It's a building-block library, not an end-to-end application, but for teams doing serious 3D ML research it saves substantial implementation time.