Automated One-Click Attendance System Using Deep Face Embeddings and Distance-Based Classification

Abstract: Traditional attendance systems are time-consuming and prone to manual errors. This paper proposes an automated one-click attendance system using deep learning-based face recognition. The system utilizes dlib's pre-trained ResNet face embedding model to convert facial images into 128-dimensional numerical vectors. Euclidean distance is applied to measure similarity between stored student embeddings and real-time classroom images. A tolerance threshold is used to classify students as present or absent. The system integrates a Streamlit-based user interface for image upload and attendance visualization. Experimental results demonstrate reliable recognition performance under controlled lighting conditions. The proposed framework provides a fast, accurate, and scalable solution for automated classroom attendance management.

Keywords: Face Recognition, Attendance System, Deep Metric Learning, Euclidean Distance, dlib, Streamlit.