3D tracking in point cloud

• Env: torch1.1; cuda10.1.214
• Dependencies: pip install -r requirement.txt
• Pointnet++: python setup.py install
• RoIAware_Pool3D: cd ./roiaware_pool3d , then python setup.py develop

• KITTI
• Pandaset
• Waymo The LIDAR_SOT is an evaluation benchmark based on the waymo validation set.

• Run python train_tracking_eval.py --data_dir="your dataset path" --dataSize=24

• Run python test_tracking.py --data_dir="your dataset path" -save_root_dir="your checkpoint path" --model="which checkpoint"
• You can access the trained model

This project is based on P2B, OpenPCDet, and LIDAR_SOT.