Seeing the Room, Not the Student: A Privacy-Preserving Computer Vision Prototype for Anonymous Aggregate Classroom Visual-Engagement Trend Monitoring from Face-Visibility and Head-Orientation Cues
Journal of Engineering · 01.10.2026 · 32 (10) · 1-23
DOI: 10.31026/j.eng.2026.10.01 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Engineering, online erschienen am 01.10.2026. Verfasst von Hamzah Asyrani Bin Sulaiman und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Camera-based classroom analytics can preserve a record of how a room behaved during a lesson, but conventionally built systems store face images that remain open to re-identification, name individual learners, and report weak visual cues as attention scores, the signal cannot support. The gap addressed here is an engineering one: neither the reported classroom-vision prototype nor the verification restricts storage to aggregate numbers, and both verify this restriction below the application layer. This paper presents the design, implementation and pilot evaluation of a prototype built to that constraint. Webcam frames are read in real time, passed through face detection and coarse head-orientation estimation, and released from memory; only numerical aggregate records reach disk, and a Streamlit dashboard renders the resulting trends. Three processing modules were compared under identical logging conditions: a Haar Cascade baseline, an OpenCV deep neural network (DNN) face detector, and a MediaPipe head-orientation module. A controlled pilot covering six predefined conditions retained 100 aggregate orientation records; the intended dominant category was recovered in 6 of 6 sessions (100%), with 29% front-facing, 21% looking left, 20% looking right, 17% looking down and 13% looking up. Of 23 DNN records, 22 carried a detection with a mean confidence of 0.890 to 1.000, all clearing the 0.60 acceptance threshold by at least 0.29, against a 4.3% zero-detection rate. A three-layer system-level audit recorded no write outside the aggregate CSV logs. The findings indicate that a research-safe aggregate visual analytics pipeline can be engineered, and its privacy boundary verified, before classroom-level deployment.
Zitieren
APA
Sulaiman, H. A. B., Abdullasim, N. B., Dolhalit, M. L. B., & Zolkifly, I. A. (2026). Seeing the Room, Not the Student: A Privacy-Preserving Computer Vision Prototype for Anonymous Aggregate Classroom Visual-Engagement Trend Monitoring from Face-Visibility and Head-Orientation Cues. Journal of Engineering, 32(10), 1-23. https://doi.org/10.31026/j.eng.2026.10.01
Vancouver
Sulaiman HAB, Abdullasim NB, Dolhalit MLB, Zolkifly IA. Seeing the Room, Not the Student: A Privacy-Preserving Computer Vision Prototype for Anonymous Aggregate Classroom Visual-Engagement Trend Monitoring from Face-Visibility and Head-Orientation Cues. Journal of Engineering. 2026;32(10):1-23. doi:10.31026/j.eng.2026.10.01
BibTeX
@article{Sulaiman2026,
author = {Sulaiman, Hamzah Asyrani Bin and Abdullasim, Nazreen Bin and Dolhalit, Mohamad Lutfi Bin and Zolkifly, Iznora Aini},
title = {Seeing the Room, Not the Student: A Privacy-Preserving Computer Vision Prototype for Anonymous Aggregate Classroom Visual-Engagement Trend Monitoring from Face-Visibility and Head-Orientation Cues},
journal = {Journal of Engineering},
year = {2026},
volume = {32},
number = {10},
pages = {1--23},
doi = {10.31026/j.eng.2026.10.01}
}
EUVIMED