Multiplet selection: a practical study of multi-faces recognition for student attendance system

Abstract

This paper presents a new approach, Multiplet Selection, for multi-faces recognition and its application in student attendance system. Instead of using a linear classifier such as SVM to classify face feature vectors, we adopt a "multiplet selection" approach such that Euclidean distances score between each identity's Anchor face [4] and a random input face are computed. Together with a pre-determined threshold parameter, this score is used for input face-identity pair association. We also develop a student attendance system based on the proposed multi-face recognition algorithm. And testing results video are available at the following URL: https://youtu.be/OZOgcw7B1YI.