What is a face recognition attendance system?
A face recognition attendance system marks students present by recognising their faces with a camera, instead of calling names, scanning fingerprints or tapping cards. Each student's face is enrolled once. After that, a camera pointed at the class finds each face, compares it with the enrolled faces and fills in the attendance register.
Calling the roll for a class of 40 eats into every lesson, and friends answer for each other. Fingerprint machines mean a queue and every child pressing the same sensor. Cards get forgotten or lent. A camera scan in the classroom avoids all three.
How does face recognition attendance work?
There are two stages, enrolment and the scan. In Premeena both run in the browser on the school's own phone, tablet or laptop.
- Enrolment turns a face into numbers. The student looks at the camera, then turns their head in one slow circle, which takes about five seconds. The software captures five angles and turns each into a face signature: a list of 128 numbers that summarises what the face looks like. A short live test confirms it.
- Only the numbers are saved. Camera frames are processed on the device and discarded. A signature is not an image, and the photo it came from is never stored.
- The scan finds faces. When the teacher starts a scan, the device downloads the signatures for that one class. A face-detection model called YuNet finds every face in each video frame.
- Each face is compared with the class. A face-recognition model called SFace turns each detected face into a 128-number signature and compares it with the class's enrolled signatures. A match must be close enough, and clearly closer than the next most similar student. A borderline match has to be seen twice.
- The register is pre-filled, not finalised. Recognised students become Present and everyone else Absent, in a draft. Students already marked Late or Excused are left alone. The teacher fixes anything wrong, then saves or finalises.
Matching against one class of a few dozen students, not the whole school, keeps it fast enough for a moving camera.