Triplet and Transformer Based Approaches to Face Recognition
Loading...
Authors
Davidson, Gabrielle
Rodriguez, Daniel
Tilley, Adam C.
Mahmood, Ausif
Issue Date
2023-03-24
Type
Other
Language
en_US
Keywords
Facial Recognition , Biometric Security , Triplet Architecture
Alternative Title
Abstract
With the increasing role of Al across all spectrums of digital technology, facial recognition is becoming an important aspect of biometric security. Deep Convolution based networks with the Triplet loss were quite successful (e.g., FaceNet) in facial recognition resulting in greater than 99% accuracy on benchmarks such as LFW. With the recent success of Transformer based Natural Language Processing architectures (e.g., ChatGPT), transformers have been attempted in Computer Vision applications and have shown considerable success with better computational efficiency as compared to CNN based architectures. In this work, we compare the FaceNet architecture and the transformer-based architecture for facial recognition and provide an insightful understanding of the facial recognition process, its limitations, and future directions.
Description
UB Rise 2023
Department of Computer Science and Engineering
School of Engineering