Exploiting Face Recognizability with Early Exit Vision Transformers

Face recognition with Deep Learning is generally approached as a problem of capacity. The field has seen progressively deeper, more complex models or larger, more highly variant datasets. However, the carbon footprint of machine learning (ML) is a concern. A real push is developing to reduce the energy consumption of ML as we strive for a more eco-friendly society. Lower energy consumption or compute budget is always desirable, if accuracy is not reduced below a usable level. We present an approach using the state of the art Vision Transformer and Early Exits for reducing compute budget without significantly affecting performance. We develop a system for face recognition and identification with a closed-set gallery and show that with a small reduction in performance, a reasonable reduction in FLOPs can be obtained using our method.