Since last summer, I have been working on a 3D LiDAR project that involves implementing sensor data acquisition, object detection, and tracking algorithms. For object detection, I used Depth Clustering, a fast and lightweight point cloud clustering algorithm. Due to the constraints of the environments where I needed to run, I ported the algorithm from the original C++ version to Python using Numba. I have also published the project to GitHub:
Numba does SIMD optimization with LLVM, so I will try to get some benchmarks in the future.