Smart Attendance System: Integrating Facial Recognition and Geofencing for Educational Institutions
Adeleke University Journal of Science · 14.08.2026 · 5 (1) · 1069-1083
DOI: 10.66945/aujs.v5i1.363 ↗Kurz gefasst
Fachartikel in der Zeitschrift Adeleke University Journal of Science, online erschienen am 14.08.2026. Verfasst von Adedamola Akintunde und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Background
Attendance in many educational institutions is still taken by roll call or paper sign-in sheet, methods that eat into teaching time, invite clerical error, and are easy to defeat through “buddy punching”, where one student signs in for an absent friend. A questionnaire we administered to 81 respondents (78 students, 3 lecturers) at Adeleke University found that 61.7% spent over thirty minutes recording attendance in classes above fifty students, and 91.4% said they would rather use a technology-based alternative.
Aim
This study set out to design, build, and evaluate a Smart Attendance System (SAS) that pairs facial recognition with geofencing, so that attendance is only accepted when a student's identity and physical location can both be confirmed.
Methods
Development ran from December 2024 to mid-2025 using an incremental approach. The system itself is a three-layer stack: a React Native mobile app, a Next.js dashboard for staff, and an ASP.NET Core Web API backend with a Microsoft SQL Server database. Identity checks were handled by the Python face recognition library at a tuned decision threshold; location checks used a point-in-polygon algorithm against stored classroom boundaries. Everything ran on a local network. We tested facial recognition across 14 unique face-pair comparisons (7 genuine, 7 impostor) at several decision thresholds, and tested geofencing along a single boundary transect (9 points, one device).
Results
The system implemented authentication, course registration, scheduling, and dual-verification sign-in with record export, all working as intended. At the face recognition library's default threshold, the false accept rate was 80.0% — far too permissive for a fraud control; tightening the threshold to 0.45 removed false accepts altogether (0.0% FAR) at the cost of a 14.3% false reject rate, while 0.50 held a 100.0% true accept rate. On the geofencing side, sensitivity (correctly accepting genuine in-boundary attempts) was 100.0%, but specificity (correctly rejecting genuine out-of-boundary attempts) was only 25.0%: the boundary accepted positions up to 5m beyond its true edge.
Conclusion
A dual-layer verification approach combining facial recognition and geofencing can feasibly automate attendance capture, but only if the facial recognition threshold is deliberately tuned rather than left at its library default. The geofencing boundary tested here turned out to be too loose rather than too tight, a more precise and more useful finding than the anecdotal reports of over-strictness noted earlier in development. Larger-sample, multi-venue testing is recommended before wider deployment. Keywords: Facial recognition; Geofencing; Smart attendance system; Proxy attendance; Educational technology
Zitieren
APA
Akintunde, A., Ibrahim, M. O., & Arowolo, E. O. (2026). Smart Attendance System: Integrating Facial Recognition and Geofencing for Educational Institutions. Adeleke University Journal of Science, 5(1), 1069-1083. https://doi.org/10.66945/aujs.v5i1.363
Vancouver
Akintunde A, Ibrahim MO, Arowolo EO. Smart Attendance System: Integrating Facial Recognition and Geofencing for Educational Institutions. Adeleke University Journal of Science. 2026;5(1):1069-1083. doi:10.66945/aujs.v5i1.363
BibTeX
@article{Akintunde2026,
author = {Akintunde, Adedamola and Ibrahim, Michael Oluwasegun and Arowolo, Emmanuel Olalekan},
title = {Smart Attendance System: Integrating Facial Recognition and Geofencing for Educational Institutions},
journal = {Adeleke University Journal of Science},
year = {2026},
volume = {5},
number = {1},
pages = {1069--1083},
doi = {10.66945/aujs.v5i1.363}
}
EUVIMED