3D object detection on LiDAR point clouds using BEVFusion and MMDetection3D. Includes custom patches for spconv 2.x and full compatibility with Python 3.10, PyTorch 2.1.2, and CUDA 11.8 on the nuScenes benchmark.
computer-vision deep-learning cuda pytorch perception lidar autonomous-driving point-clouds 3d-lidar google-colab 3d-object-detection computer-vision-ai nuscenes birds-eye-view mmdetection3d bevfusion mmdetection3d-fork
-
Updated
Nov 9, 2026 - Jupyter Notebook