qiaoliuhit / mmnet Goto Github PK

We propose a feature model comprising TIR-specific discriminative features and fine-grained correlation features for TIR object representation. Then, we develop a multi-task matching framework (MMNet) to integrate these two features for robust TIR tracking. In addition, we build a large-scale TIR training dataset to train the network for adapting the model to the TIR domain. More details please see our paper, supplementary material.

• You can download the proposed TIR training dataset from Baidu Pan. [News 2020-08] We have extended this dataset to a new TIR object tracking benchmark, LSOTB-TIR.
• You can download several our trained models from Baidu Pan or MEGA drive.
• You can download the tracking raw results of three benchmarks from Baidu Pan or MEGA drive.
• We provide a raw result of MMNet on the LSOTB-TIR Benchmark in here.

1. Clone the code and unzip it in your computer.
2. Prerequisites: Ubuntu 14, Matlab R2017a, GTX1080, CUDA8.0.
3. Download our trained models from here and put them into the src/tracking/networks folder .
4. Run the run_demo.m in src/tracking folder to test a TIR sequence using a default model.
5. Test other TIR sequences, please download the PTB-TIR dataset from here.

1. Preparing your training data like that in here. Noting that preparing the TIR training data uses the same format and method as the above.
2. Configure the path of training data in src/training/env_path_training.m.
3. Run src/training/run_experiment_MMNet.m. to train the proposed MMNet.
4. The network architecture and trained models are saved in src/training/data-MMNet folder.

If you use the code or dataset, please consider citing our paper.

@inproceedings{MMNet,
  title={Multi-Task Driven Feature Models for Thermal Infrared Tracking},
  author={Liu, Qiao and Li, Xin and He, Zhenyu and Fan, Nana and Yuan, Di and Liu, Wei and Liang, YongSheng},
  booktitle={Thirty-Fourth AAAI Conference on Artificial Intelligence},
  pages={11604-11611},
  year={2020}
}
or
The extened journal version
@article{MMNet,
  title={Learning Dual-Level Deep Representation for Thermal Infrared Tracking},
  author={Liu, Qiao and Yuan, Di and Fan, Nana and Gao, Peng and Li, Xin and He, Zhenyu},
  journal={IEEE Transactions on Multimedia},
  year={2022}
}

Feedbacks and comments are welcome! Feel free to contact us via [email protected] or [email protected]