Conventional classroom attendance systems are ineffective, prone to errors, and fail to provide insights into student engagement or behavioral dynamics. This paper talks about AI-Based Smart Classroom Attendance System with Engagement Analytics (ASCAS), a full-stack, ready for production app that uses deep learning to recognize faces, analyze emotions in real time, find phones, and create role-based multi-user dashboards to automate attendance marking while also measuring how engaged students and teachers are. The system uses a hybrid face recognition pipeline that combines InsightFace (ArcFace model) for 512-dimensional deep embedding extraction and Local Binary Pattern Histograms (LBPH) for texture-based fallback verification. It can recognize faces 80% accuracy on standard laptop hardware (Intel R620 integrated GPU) and 90–95% accuracy on dedicated GPU workstations. There are three types of users that can use it: There are three types of people: students, teachers, and guests. Students and teachers sign up through a webcam-based registration portal that records facial embeddings in a MySQL relational database.
Attendance is marked automatically here as Present or Absent based on how visible the face is during configurable session times. A Convolutional Neural Network (CNN) trained on FER2013 classifies seven emotion categories for each detected face frame. The emotion scores, along with data on how long each emotion was present, create role-based engagement metrics that are shown on real-time React dashboards. A static threshold-based phone detection module sends out live alerts when it sees someone using a mobile device in the classroom. Registered stakeholders automatically get session reports by email, and they can also be exported as CSV files. An evaluation of the system on standard laptop hardware confirms 80% an engagement-score Pearson correlation of face recognition accuracy
