CuVoxelization is an CUDA implementation of voxelization. The given points is a two-dimension torch.Tensor # [num_points, 3], and the output voxels is a three-dimension torch.Tensor # [num_voxels, max_points, 3].

The enviroment of my developer machine:

• Python 3.8.8+
• PyTorch 1.10.2
• CUDA 11.1

python setup.py install

Or use:

pip install .

Or use:

pip install https://github.com/lzhnb/CuVoxelization

• Examples (More Example)
• Optimize the code
• More elegant Python Wrapper
• Support backward
• Visualization

Put the points.npy file under examples directory, then run

python verify.py