Developing Imperceptible Adversarial Patches to Camouflage Military Assets From Computer Vision Enabled Technologies

How to use:

Clone the repository:

git clone https://github.com/ChrisWise07/imperceptible-patch-generator.git

Move into the root directory:

cd <path-to-repo>/imperceptible-patch-generator

Create a virtual environment (optional):

python3 -m venv ./venv

Activate the virtual environment (optional):

source venv/bin/activate

Install required files:

pip install -r requirements.txt

Run:

bash ./execute_imperceptible_patch.sh

View Results:

cd ./code_and_experiment_data/experiment_data/<experiment-name>

View the hyper-parameters:

To see the list of hyper-parameters and default values, run:

python ./code_and_experiment_data/main.py --help

Modify the hyper-parameters:

To modify the hyper-parameters, open the file <path-to-repo>/imperceptible-patch-generator/execute_imperceptible_patch.sh and modify by adding the following to the file:

--<hyper-parameter-name> <hyper-parameter-value>

Credit

The code has adapted two libraries:
https://github.com/Trusted-AI/adversarial-robustness-toolbox
https://github.com/ZhengyuZhao/PerC-Adversarial