Ad^2Attackļ¼šAdaptive Adversarial Attack on Real-Time UAV Tracking

Demo video

  • ?
    Our video on bilibili demonstrates the test results of Ad^2Attack on several sequences.

Ad^2Attack

Environment setup

This code has been tested on Ubuntu 18.04, Python 3.8.3, Pytorch 0.7.0/1.6.0, CUDA 10.2. Please install related libraries before running this code:

pip install -r requirements.txt

Attack on Trackers

[SiamAPN]

The pre-trained model of SiamAPN can be found at (epoch=37) : general_model(code:w3u5) and the pre-trained model of Ad^2Attack can be found at /checkpoints/AdATTACK/model.pth

Ad^2Attack on other trackers, e.g., SiamCAR, SiamGAT, HiFT, SiamAPN++ will be released soon.

Datasets Setting

We evaluate our attack method on 3 well-known UAV tracking benchmark, i.e., UAV123, UAV112 and UAVDT You can download them and put them in /pysot/test_dataset remember change the path in Setting.py

Test Attack

vim ~/.bashrc
export PYTHONPATH=/home/user/Ad^2Attack:$PYTHONPATH
export PYTHONPATH=/home/user/Ad^2Attack/pysot:$PYTHONPATH
export PYTHONPATH=/home/user/Ad^2Attack/pix2pix:$PYTHONPATH
source ~/.bashrc


python pysot/tools/test.py 	        \
	--trackername SiamAPN           \ # tracker_name
	--dataset V4RFlight112          \ # dataset_name
	--snapshot snapshot/general_model.pth   # model_path

The testing result will be saved in the results/dataset_name/tracker_name directory.

Evaluation

If you want to evaluate the Ad^2Attack on trackers, please put those results into results directory.

python pysot/tools/eval.py 	                          \
	--tracker_path ./results          \ # result path
	--dataset V4RFlight112            \ # dataset_name
	--tracker_prefix 'general_model'  \ # tracker_name

Contact

If you have any questions, please contact me.

Sihang Li

Email: [email protected]

Acknowledgement

The code is implemented based on pysot, and CSA. We would like to express our sincere thanks to the contributors.

GitHub

https://github.com/vision4robotics/Ad2Attack