The code mainly requires the following libraries and you can check requirements.txt for more environment requirements.

• PyTorch 1.7.0/1.7.1
• Cuda 11.0/11.1
• pytorch3d 0.6.0
• open3d 0.13.0

Please run the following commands to install point_utils

cd model/PointUtils
python setup.py install

Training device: NVIDIA RTX 3090

The point cloud pairs list and the ground truth relative transformation are stored in data/kitti_list, data/nuscenes_list and data/apollo_list. The data of the three datasets should be organized as follows:

DATA_ROOT
├── 00
│   ├── velodyne
│   ├── calib.txt
├── 01
├── ...

DATA_ROOT
├── v1.0-trainval
│   ├── maps
│   ├── samples
│   │   ├──LIDAR_TOP
│   ├── sweeps
│   ├── v1.0-trainval
├── v1.0-test
│   ├── maps
│   ├── samples
│   │   ├──LIDAR_TOP
│   ├── sweeps
│   ├── v1.0-test

DATA_ROOT
├── TrainData
│   ├── BaylandsToSeafood
│   ├── ColumbiaPark
│   ├── ...
├── TestData
│   ├── BaylandsToSeafood
│   ├── ColumbiaPark
│   ├── ...

We provide training scripts in scripts/.

Please specify the following entries:

• DATASET: ['kitti','nusc','apollo']
• ROOT: Root of the dataset
• DATA_LIST: Data list in data/data_list, e.g., data/data_list/kitti_list
• CKPT_DIR: The dir you want to save the ckpt and log files
• NPOINTS: 16384 for kitti and apollo, 8192 for nuscenes
• pretrain_feats: Pretrain weights for feature extractor
• GPU: GPU Id if you have multiple GPUs

We provide pre-trained weights for three datasets in ckpt/pretrained/kitti_release/, ckpt/pretrained/nusc_release/ and ckpt/pretrained/apollo_release/, respectively. And the test scripts are provided in scripts/.

Please specify the following entries:

• DATASET: ['kitti','nusc','apollo']
• ROOT: Root of the dataset
• DATA_LIST: Data list in data/data_list, e.g., data/data_list/kitti_list
• SAVE_DIR: The dir you want to save the results
• PRETRAIN_WEIGHTS: Pretrain weights in ckpt/pretrained, e.g., ckpt/pretrained/kitti_release/kitti.pth
• NPOINTS: 16384 for kitti and apollo, 8192 for nuscenes