Implementation of papers in 100 lines of code.
Metal-native differentiable 3D Gaussian Splatting for Apple Silicon: fused forward+backward rasterizer on PyTorch MPS, no CUDA anywhere.
Implementation of papers in 100 lines of code.
Metal-native differentiable 3D Gaussian Splatting for Apple Silicon: fused forward+backward rasterizer on PyTorch MPS, no CUDA anywhere.