fpthink / IFP

Facilitating 3D Object Tracking in Point Clouds with Image Semantics and Geometry

Introduction

This repository is released for IFP (Image Fuse PointCloud) in our PRCV 2021 paper. Here we include our IFP model (PyTorch) and code for data preparation, training and testing on KITTI tracking dataset.

Preliminary

• conda

    conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.0

• Install dependencies.

    pip install -r requirements.txt

• Build _ext module.

   cd lib/pointops && python setup.py install && cd ../../
   cd ./pointnet2/utils/DCNv2 && python setup.py build develop

• Download the dataset from KITTI Tracking.

  Download velodyne, calib and label_02 in the dataset and place them under the same parent folder.

Evaluation

Train a new P2B model on KITTI data:

python train_tracking.py --data_dir=<data path> 

Test model on KITTI data:

python test_tracking.py --data_dir=<data path> 

Please refer to the code for setting of other optional arguments, including data split, training and testing parameters, etc.

If you think it is a useful work, please consider citing it.

@inproceedings{wang2021IFP,
  title={Facilitating 3D Object Tracking in Point Clouds with Image Semantics and Geometry},
  author={Lingpeng, Wang and Le, Hui and Jin, Xie},
  booktitle={PRCV},
  year={2021}
}

Acknowledgements

Thank Qi for his implementation of P2B.