Skip to content

Latest commit

 

History

History
18 lines (13 loc) · 496 Bytes

File metadata and controls

18 lines (13 loc) · 496 Bytes

Adversarial robustness

Download the data

mkdir -p data
curl -X GET https://s3.amazonaws.com/fast-ai-imageclas/oxford-iiit-pet.tgz data/pets.tgz
tar -xzf data/pets.tgz -C data

Then run

poetry run python ch08/adversarial/main.py

to train a neural network model for image classification, perform adversarial attacks using the Fast Gradient Sign Method (FGSM), and evaluate the model's performance using both standard predictions and Monte Carlo dropout.