Face recognition

Optimizing the design of semi-supervised facial recognition framework

• Developed an end-to-end framework for practical facial recognition using state of the art FaceNet embeddings. The framework allowed for quick semi-supervised face data labelling, training, deployment and performance validation all through a single web-app.
• Proposed a novel clustering algorithm for semi-supervised data labelling and a K-Means based classifier adhered to enhance performance in the classroom environment.
• Gathered a face dataset in lecture halls, first such dataset in real-world settings to be the best of our knowledge.
• Demonstrated the utility of framework by automating attendance in 3 classrooms at IIT Tirupati with a training and deployment time of just a few minutes for any new classroom.

Full Report Web application demo video Paper Overleaf - A note on recognition accuracy Github - Application code