Awesome
SiameseTracking4UAV
Siamese Object Tracking for Unmanned Aerial Vehicle: A Review and Comprehensive Analysis
Changhong Fu, Kunhan Lu, Guangze Zheng, Junjie Ye, Ziang Cao, Bowen Li, and Geng Lu
This work has been accepted and published by Artificial Intelligence Review (JCR Q1, IF = 12).
This code library gives our experimental results and most of the publicly available Siamese trackers.
The trackers are in folder experiments and the results are in results.
Paper link: https://link.springer.com/article/10.1007/s10462-023-10558-5
View-only link: https://rdcu.be/dhWgD
If you want to use our code libary, experimental results, and related contents, please cite our paper using the format as follows:
@article{Fu2023SiameseOT,
title={{Siamese Object Tracking for Unmanned Aerial Vehicle: A Review and Comprehensive Analysis}},
author={Fu, Changhong and Lu, Kunhan and Zheng, Guangze and Ye, Junjie and Cao, Ziang and Li, Bowen and Lu, Geng},
journal={Artificial Intelligence Review},
volume={56},
number={Suppl 1},
pages={1417-1477},
year={2023},
doi={10.1007/s10462-023-10558-5}
}
Results_OPE_AGX
The trackers are tested on the following platform.
- Ubuntu 18.04
- 8-core Carmel ARM v8.2 64-bit CPU
- 512-core Volta GPU
- 32G RAM
- CUDA 10.2
- Python 3.6.3
- Pytorch 0.7.0/1.6.0
All the Siamese trackers' results are obtained using an NVIDIA Jetson AGX Xavier.
Figures
Here shows some of the tracking results of 19 SOTA Siamese trackers.
Comparison of the performance under all the six authoritative UAV tracking benchmarks. <img src="./figures/Precision and FPS.png">
<img src="./figures/Normalized Precision and FPS.png"> <img src="./figures/Success and FPS.png">The average performance comparison of the five real-time Siamese trackers under all the six authoritative UAV tracking benchmarks. <img src="./figures/Attributes.png">
Environment setup
This code has been tested on an NVIDIA Jetson AGX Xavier with Ubuntu 18.04, Python 3.6.3, Pytorch 0.7.0/1.6.0, CUDA 10.2.
Please install related libraries before running this code:
pip install pyyaml yacs tqdm colorama matplotlib cython tensorboardX easydict
Test
Download pretrained models form the links in experiments
directory or download pretrained models from official code site and put them into experiments
directory.
Download testing datasets and put them into test_dataset
directory. If you want to test the tracker on a new dataset, please refer to pysot-toolkit to set test_dataset.
The papers whose benchmarks are used in the experimental evaluations are listed here.
UAV123@10fps & UAV20L
Paper: A Benchmark and Simulator for UAV Tracking
Paper site: https://link.springer.com/chapter/10.1007%2F978-3-319-46448-0_27 .
Code and benchmark site: https://cemse.kaust.edu.sa/ivul/uav123 .
DTB70
Paper: Visual Object Tracking for Unmanned Aerial Vehicles: A Benchmark and New Motion Models
Paper site: https://dl.acm.org/doi/10.5555/3298023.3298169 .
Code and benchmark site: https://github.com/flyers/drone-tracking .
UAVDT
Paper: The Unmanned Aerial Vehicle Benchmark: Object Detection and Tracking
Paper site: https://link.springer.com/article/10.1007/s11263-019-01266-1 .
Code and benchmark site: https://sites.google.com/site/daviddo0323/projects/uavdt .
VisDrone-SOT2020
Paper: VisDrone-SOT2020: The Vision Meets Drone Single Object Tracking Challenge Results
Paper site: https://link.springer.com/chapter/10.1007/978-3-030-66823-5_44 .
Code and benchmark site: http://aiskyeye.com/ .
UAVTrack112
Paper: Onboard Real-Time Aerial Tracking With Efficient Siamese Anchor Proposal Network
Paper site: https://ieeexplore.ieee.org/abstract/document/9477413 .
Code and benchmark site: https://github.com/vision4robotics/SiamAPN .
UAVDark135
Paper: All-Day Object Tracking for Unmanned Aerial Vehicle
Paper site: https://ieeexplore.ieee.org/document/9744417 .
Code and benchmark site: https://github.com/vision4robotics/ADTrack_v2 .
UAVDarkTrack2021
Paper: Tracker Meets Night: A Transformer Enhancer for UAV Tracking
Paper site: https://ieeexplore.ieee.org/document/9696362 .
Code and benchmark site: https://github.com/vision4robotics/SCT .
NAT2021
Paper: Unsupervised Domain Adaptation for Nighttime Aerial Tracking
Paper site: https://ieeexplore.ieee.org/document/9879981 .
Code and benchmark site: https://vision4robotics.github.io/NAT2021/ .
Option 1
Use the corresponding 'tools/test_<tracker_name>.py' to test the performance of the tracker. Take the test of SiamAPN as an example:
python tools/test_siamapn.py \
--dataset UAVTrack112 \ # dataset_name
--datasetpath ./test_dataset \ # dataset_path
--config ./experiments/SiamAPN/config.yaml \ # tracker_config
--snapshot ./experiments/SiamAPN/model.pth \ # tracker_model
--trackername SiamAPN # tracker_name
The testing result will be saved in the results/<dataset_name>/<tracker_name>
directory.
The settings required by different trackers will be different. For details, please refer to the examples in 'tools/test.sh'
Option 2
Similar to Option 1, a more convenient way of testing is provided using 'tools/test.py' to test all the trackers.
python tools/test.py \
--dataset UAVTrack112 \ # dataset_name
--datasetpath ./test_dataset \ # dataset_path
--config ./experiments/SiamAPN/config.yaml \ # tracker_config
--snapshot ./experiments/SiamAPN/model.pth \ # tracker_model
--trackername SiamAPN # tracker_name
Evaluation
If you want to evaluate the trackers mentioned above, please put those results into results
directory as results/<dataset_name>/<tracker_name>
.
python tools/eval.py \
--dataset UAVTrack112 \ # dataset_name
--datasetpath path/of/your/dataset \ # dataset_path
--tracker_path ./results \ # result_path
--tracker_prefix 'SiamAPN' # tracker_name
Contact
If you have any questions, please contact me.
Kunhan Lu
Email: lukunhan@tongji.edu.cn .
Acknowledgements
-
The code is implemented based on pysot. We would like to express our sincere thanks to the contributors.
-
We would like to thank Jilin Zhao, Kunhui Chen, Haobo Zuo, and Sihang Li for their help in building this code library.
-
We also thank the contribution of Matthias Muller, Siyi Li, Dawei Du, Heng Fan et al. for their previous work of the benchmarks UAV123@10fps, UAV20L, DTB70, UAVDT, and VisDrone-SOT2020-test.