• Trained a pre-trained resnet18, with its weights frozen, added a classifier, and finetuned the classifier layers Training_CelebA.ipynb

• Evaluated on the test data and obtained accuracies wrt every attribute inference.py, attr acc.png

• Based on some analysis of the data population and the above results, chose "Male" as the spurious attribute and "Bald" as the target (affecting) attribute.

• Then trained a Binary Classifier (a finetuned resnet18) and observed the bias (Equalized Odds). Bald_Classifier(Biased).ipynb, equal_odds.py.

• Implemented an adversarial training pipeline to reduce the above bias. adversarial_training.ipynb

• Lastly, rechecked the bias to ensure Equalized Odds is satisfied. recheck_odds.py

• Conclusion: Adversarial Training helped reduce the original bias by 20%