Work

Discovering Interpretable Feature Directions in the Embedding Space of Face Recognition Models