Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{198350,
        author = {Prathamesh Sankpal and Atharva Mahangade and Nikhil Jakapure and Samarth Telkar},
        title = {Optimizing Euclidean Distance Thresholds in Deep Facial Embeddings for High-Accuracy Real-Time Attendance Tracking},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {8399-8401},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=198350},
        abstract = {The automated system for tracking presence through computer vision has become one of the most important innovations for the problems faced by many schools and offices. Although deep learning models produce very accurate facial embeddings, their use in the real world demands careful optimization of classification thresh-olds to guarantee high levels of both accuracy and convenience. This paper introduces the end-to-end automated system for attendance tracking and aims at optimizing Euclidean distance thresholds within deep facial embeddings. Using CNNs with ResNet-34 architecture, the proposed approach computes the 128-dimensional facial embeddings of live webcam images and compares them to a local database. At the core of this work, the evaluation of the proposed system through various Euclidean distance thresholds (0.4, 0.6, and 0.8) is performed, and the effects of those changes on True Positive and False Positive rates are analyzed. Through this process, the mathematically optimal threshold is found in terms of minimizing False Negatives without producing any false-positive outcomes. Lastly, the implementation of the developed machine learning model is showcased in the context of real-time Streamlit web-based dashboard, allowing live calculations of the system's status. (” Late” vs.” On Time”) and automated SMTP reporting.},
        keywords = {Biometric tracking, convolutional neural networks, deep facial embeddings, edge computing, Euclidean distance, real-time systems.},
        month = {April},
        }