Original Repository: open-mmlab/OpenPCDet
pclpy Repository: davidcaron/pclpy

---

OpenPCDet Installation

---

Change Record

1. Add a file convert_pcd2bin.py
   📌 need pclpy
2. pcdet/datasets/kitti/kitti_dataset.py
3. tools/cfgs/kitti_models/voxel_rcnn_car.yaml
4. tools/cfgs/dataset_configs/kitti_dataset.yaml
5. Add a file match_label.py
   • label이 없는 calib, image, velodyne 파일 삭제

---

Commands

• Convert pcd files to bin files

$ conda activate pclpy
$ python convert_pcd2bin.py
$ conda deactivate

• Generate the data infos
  kitti_dataset.yaml에서 DATA_PATH 변경해주기!

$ conda activate OpenPCDet
$ cd ~/OpenPCDet
$ python -m pcdet.datasets.kitti.kitti_dataset create_kitti_infos tools/cfgs/dataset_configs/kitti_dataset.yaml

• Train with a single GPU
  voxel_rcnn_car.yaml, pv_rcnn.yaml 사용
  

$ cd ~/OpenPCDet/tools
$ python train.py --cfg_file cfgs/kitti_models/voxel_rcnn_car.yaml --batch_size 1 --epochs 10000 

• Run the demo with the trained model and custom point cloud data

$ cd ~/OpenPCDet/tools
$ python demo.py --cfg_file cfgs/kitti_models/voxel_rcnn_car.yaml \
    --ckpt pv_rcnn_8369.pth \
    --data_path ${POINT_CLOUD_DATA}

• Test and evaluate the trained models
  • To test the specific checkpoint, use --ckpt ${CKPT}
  • To test all the saved checkpoints, use --eval_all

$ python test.py --cfg_file ${CONFIG_FILE} --batch_size ${BATCH_SIZE} --ckpt ${CKPT}
$ python test.py --cfg_file ${CONFIG_FILE} --batch_size ${BATCH_SIZE} --eval_all

<주의사항>

nas에 있는 파일을 불러오는 경우 permission denied 에러 뜰 수 있음.
이땐 conda activate ${환경이름}후에 python 명령어 쓰지 말고, sudo ~/anaconda3/envs/${환경이름}/bin/python 뒤에 명령어 쓰기
ex) python convert_pcd2bin.py 대신 sudo ~/anaconda3/envs/pclpy/bin/python convert_pcd2bin.py