README.md

Train credential detection model

This code will allow you to retrain model on the CredData dataset

Preparation

• Make sure that you are using Python 3.7.10 or higher

• Download CredData dataset

git clone https://github.com/Samsung/CredData
cd CredData
pip install PyYAML
python download_data.py --data_dir data

• Go back to CredSweeper/experiment directory
• Install the requirements

pip install -r requirements.txt

• Make sure that credsweeper in the PYTHONPATH. You can add it with

export PYTHONPATH=<CredSweeper directory>:$PYTHONPATH

Example:

export PYTHONPATH=/home/user/code/CredSweeper:$PYTHONPATH

Run

• Launch the experiment with

python main.py --data <CredData location> -j <num parallel process to run>

Example:

python main.py --data /home/user/datasets/CredData -j 16

• Resulting model will be saved to results/ml_model_at-<date_time>. You now can convert the model to onnx:

python -m tf2onnx.convert --saved-model results/ml_model_at-20240225_111951 --output ../credsweeper/ml_model/ml_model.onnx --verbose