Attendance Checking-Based On-Device Multiple Face Recognition using Flutter Platform
Pitchayapatchaya Srikram, Patamavadee Tubthong, Siripassorn Meewawsang, Sumonjit Sangkhao, Jedsada Arunruerk · 2024
Biometric class attendance systems have recently been developed to increase accuracy and efficiency, with facial recognition a popular choice. Currently, facial recognition mobile applications can recognize one student per frame, which is not appropriate for classes with vast numbers of students. This paper proposed a student attendance mobile application implemented on the Flutter platform. Facial detection was achieved using the Firebase ML kit, with multiple-facial recognition performed by a deep neural network (DNN) model that applied transfer learning using the fine-tuned ResNet-50 and VGGFace2 datasets.Facial images of 25 students were assessed, with results giving 0.95 precision, 0.875 recall, 0.91 F-measure, 0.91 accuracy, and 24.5 seconds processing time. The proposed mobile application was then deployed as a smaller model, giving precision, recall, F-measure, and accuracy similar to the baseline model but with reduced computational time.