Implemented CNNs and DB-VAE for bias mitigation in facial detection systems.
Developed CNN architectures for facial detection and integrated a Debiasing Variational Autoencoder (DB-VAE) to reweight training samples. Improved fairness by increasing sampling for underrepresented facial features, reducing demographic bias in model predictions.
