Face Recognition in Chatbot-Based Attendance System with Multi-Task Cascaded Convolutional Neural Network (MTCNN) and GhostFaceNetV2

The attendance process is a very important process in various activities, one of which is in lecture activities. Attendance will affect the evaluation of activities. The manual attendance process can trigger cheating and various other problems. The current attendance system is very diverse, such as online attendance, fingerprint use, and facial recognition. The use of facial recognition technology has become a major concern in the development of artificial intelligence-based attendance solutions. This study aims to implement facial recognition technology in an attendance system integrated with a chatbot using the MTCNN algorithm and GhostFaceNetV2 architecture. The CRISP-DM method is used in the development stages of this system, starting from understanding needs, data processing, modeling, evaluation, to sending chatbots to the WhatsApp platform, which is the distribution stage in this study. The final result of this study is a WhatsApp chatbot built using Node.js, Flask, and Strapi, which is capable of performing facial recognition-based attendance with an accuracy of 94.89%, with a detection time using a CPU of 1.78 seconds with a CPU increase of 45.7% and an increase in RAM of 0.1%. This study shows that the application of facial recognition technology in a chatbot-based attendance system can increase efficiency and accuracy in the attendance process.