Face Recognition Using ArcFace and FaceNet in Google Cloud Platform For Attendance System Mobile Application
Rosa Andrie Asmara, Brian Sayudha, Mustika Mentari, Rizky Putra Pradhana Budiman, Anik Nur Handayani, Muhammad Kholil Ridwan, Putra Prima Arhandi · 2022
The attendance system process in Indonesia generally are still using a traditional method.Paper-based is used as a medium to perform attendance at every event.With this traditional method, there are still many shortcomings in terms of security and management.In terms of security, the traditional attendance system is still quite lacking due to the number of participants cheating by asking their relatives, such as examples of signatures that can still be imitated, or attendance checks can still be tricked because we can change them easily.Therefore, it is necessary to have an attendance system that can be carried out efficiently, safely, and easy to manage, with attendance being done online or using a smartphone.It can be implemented easier for event owners to manage the attendance track of participants, reduce the use of paper, which is quite significant, and secure the attendance system.CNN is an artificial neural network that is more often used in visual image analysis.CNN can distinguish visual images from one another with various aspects given.The models that we used for this application are ArcFace and FaceNet.Three different BackEnd Encoder and BackEnd recognized are used, RetinaFace, MTCNN, and OpenCV.From the experiment, we suggest the usage of ArcFace in RetinaFace for High Accuracy of recognition but with high-cost drawbacks, the longer computation time for encoding and recognition.As an alternative, ArcFace with MTCNN can be used with faster computation time but less accurately than RetinaFace.