DS482 Face Recognition - Adversial Examples as Privacy Defenses

The actual AI system can be found in the vgg_face2_transfer_learning notebook.
The audit part can be found in the vgg_face2_adversarial_examples notebook.
The cloaking process can be found in the cloak notebook
extract_faces.py takes the images from the original Kaggle dataset (linked in the notebook) and extracts the face from the image. These extraced faces for all 17 celebrities are stored in the celebrity_faces_cropped_dataset folder
The folder vgg_face_17_celebs contains the saved model after re-training
The folder vgg_face_adversarial contains the saved model after re-training with adversarial examples

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
data		data
test_images		test_images
vgg_face_17_celebs		vgg_face_17_celebs
vgg_face_adversarial		vgg_face_adversarial
README.md		README.md
cloak.ipynb		cloak.ipynb
extract_faces.py		extract_faces.py
preprocess.ipynb		preprocess.ipynb
vgg_face2_adversarial_examples.ipynb		vgg_face2_adversarial_examples.ipynb
vgg_face2_transfer_learning.ipynb		vgg_face2_transfer_learning.ipynb

Provide feedback