Simple 3D Packing
4 months ago
- #GPU-acceleration
- #FFT
- #3D-packing
- 348 objects from Thingi10K packed into a 240×123×100mm tray at 60.8% density.
- GPU-accelerated 3D bin packing using FFT-based collision detection.
- Implements the spectral packing algorithm from SIGGRAPH 2023.
- Features include GPU-acceleration, Python API, and Blender export.
- Requires Python 3.8+, CUDA 11.0+, and NVIDIA GPU with compute capability 6.0+.
- Installation involves cloning the repository and installing dependencies.
- Includes a simple Python API for packing voxelized items or STL files.
- Algorithm workflow: voxelization, collision detection, scoring, and greedy placement.
- Examples provided for basic packing, benchmarking, and generating teaser images.
- Project structure includes C++/CUDA implementation and Python bindings.
- MIT License and citation request for the original SIGGRAPH 2023 paper.