Awesome

Multi-Correlation Siamese Transformer Network with Dense Connection for 3D Single Object Tracking

Introduction

This repository is the official implementation of our paper "Multi-Correlation Siamese Transformer Network with Dense Connection for 3D Single Object Tracking, IEEE Robotics and Automation Letters, 2023". Our code is based on V2B and SST.

Environment settings

• Create an environment for MCSTN

conda create -n MCSTN python=3.7
conda activate MCSTN

• Install pytorch and torchvision

conda install pytorch==1.9.1 torchvision==0.10.1 torchaudio==0.9.1 cudatoolkit=10.2 -c pytorch

• Install dependencies.

pip install -r requirements.txt

Data preparation

We use the datasets KITTI, nuScenes, and Waymo in the same way as V2B and STNet. Please refer to V2B for the detail of data preparation.

Note: After you get the dataset ready, please modify the path variable data_dir and val_data_dir about the dataset under configuration file MCSTN_main/utils/options.

Evaluation

Train a new model:

python main.py --which_dataset KITTI/NUSCENES --category_name category_name

Test a model:

python main.py --which_dataset KITTI/NUSCENES/WAYMO --category_name category_name --train_test test

Please refer to the relevant code for more details of the parameter setting.

Visualization

cd MCSTN_main/visualization/
python visual.py

Citation

If you find the code or trained models useful, please consider citing:

@ARTICLE{mcstn2023,
  author={Feng, Shihao and Liang, Pengpeng and Gao, Jin and Cheng, Erkang},
  journal={IEEE Robotics and Automation Letters}, 
  title={Multi-Correlation Siamese Transformer Network With Dense Connection for 3D Single Object Tracking}, 
  year={2023},
  volume={8},
  number={12},
  pages={8066-8073},
  doi={10.1109/LRA.2023.3325715}}

Acknowledgements

• Thank Le Hui et al. for their implementation of V2B.
• Thank Ziqi Pang et al. for the Waymo 3D-SOT benchmark.