Attention Educosmos: A Smart Classroom Focus and Smartphone Monitoring System

Manual roll-calls consume teaching time and are easy to falsify, while a single teacher cannot watch a whole room for smartphone use. Existing face-recognition attendance systems stop at attendance and take no account of classroom behaviour. This paper presents Attention Educosmos, a CPU-only system that identifies students from a classroom video stream, marks attendance automatically, detects mobile phones, attributes each phone to a specific student, and responds with a graduated policy: two private warning e-mails to the student, followed by escalation to the teacher and parent only if use continues. Monitoring is bound to a configurable timetable and is switched off during breaks. The pipeline combines the YuNet face detector, dlib 128-dimensional face embeddings and YOLOv8n. We describe four design decisions that were needed to make these off-the-shelf models behave reliably together: track-level identity persistence, a zoomed per-student region-of-interest pass for small phones, a face-size-scaled ownership rule, and a timer-driven escalation state machine. We report the parameters and the measurements behind them from a single-classroom recording, and we state plainly that a formal precision/recall evaluation on a larger dataset remains future work.