Rust bindings for CUDA point cloud processing with NVIDIA cuPCL.
** This is a work in progress. **
- Voxel Downsample (libcudafilter.so) - See #39
- Passthrough (own implementation)
- Clustering (libcudacluster.so)
- ICP
- NDT
- Octree
- Segmentation
Install cuPCL from NVIDIA for your architecture.
git clone -b x86_64_lib|jp4.x|jp5.x https://github.com/NVIDIA-AI-IOT/cuPCL
sudo find cuPCL -name '*.so' -exec cp {} /usr/local/lib \;
sudo ldconfig
Check if the everything works before using it in your project.
cargo test # cpu
cargo test --no-default-features --features cuda
Use the cuda
feature for acceleration:
[dependencies]
cupcl = { git = "https://github.com/stelzo/cupcl-rs", branch = "main", default-features = false, features = "cuda" }