Auto Attendance Checker
AI attendance for every lesson
A camera above the classroom door recognises students as they come in and go out. When the lesson ends, the teacher already has the journal: who came, when each student left and how much of the lesson they attended. The teacher does nothing extra.
- Universities
- Schools and colleges
- Training centres
- frames per second, processed in real time
- 20-30
- automated tests passed
- 65/65
- trial lessons, 3 to 7 October 2026
- 18
- dependence on the internet or cloud
- 0
Inside now 0
A signature sheet only shows one moment.
The same five steps in every lesson
- 1
Sensor
Notices that someone is in the room
- 2
Gateway
Collects the signal for its floor
- 3
Server
Wakes the camera for that room
- 4
Recognition
Identifies students by face and appearance
- 5
Journal
Entry, exit and attendance share
The camera does not record all the time. It starts only on a signal from the sensor or the timetable. That protects privacy and keeps the server load low.
How the system decides who is who
By face
As a student walks through the door, their face is compared with the photos on record.
By appearance
When the face is not visible, the system uses body images collected during that lesson. This reference set is rebuilt for every lesson.
Decision rule
The face always outranks appearance. A new unknown person is never taken for a student just because of similar clothes.
Exit
Exits are recorded only in the door zone. A student who drops out of view inside the room still counts as present.
Our core rule: “Unknown” beats a wrong name
When the system is not sure, it writes “Unknown” instead of guessing. That keeps the risk of marking the wrong student as low as possible.
Biometric data never leaves the university
A face image is specially protected personal data, so privacy was part of the design from the first day.
Only on your own server
Images, templates and journals stay at the university. No cloud, no internet.
Numbers, not photos
The face database stores a numeric template derived from the image, not the image itself.
Body images for one lesson
They are used for the current lesson only and moved to the archive when it ends.
No constant video
The camera runs only when needed. There is no continuous recording.
Protected passwords
Camera passwords are never stored in plain text in program files.
Before wide rollout, students are informed, written consent is collected, and retention periods and access rights are defined. These steps are confirmed with legal advice under Azerbaijan's Law on Personal Data.
From hardware to reports, one complete solution
- Developed by our team
- Proven hardware, installed by us
- 01
Sensor system (sensor and gateway)
Detects people in the room and passes the signal, floor by floor, to the server.
- 02
IP camera (PoE)
Watches entries and exits from above the door. Power and video run over one cable.
- 03
Network
Floor, building and central switches, plus cabling.
- 04
GPU servers
Run the recognition software and store all the data.
- 05
Uninterruptible power (UPS)
Protects the servers while power switches over to the generator.
- 06
Recognition software
Face, body and person tracking, decision rules and the entry and exit journal.
- 07
Student enrolment
Each student is added with one or two reference photos.
- 08
Timetable control
Attendance starts and stops automatically with the lesson schedule.
- 09
Teacher and admin panel
Live “Inside now” view, journal, manual corrections and access control.
- 10
Reports
Attendance by group, subject and student.
- 11
Integration with university systems
Student lists come in, results go out.
- 12
Archive and backup
Journals and archives are stored safely.
- 13
Device monitoring
Keeps track of whether sensors, gateways and cameras are working.
The sensor and gateway are made by us
- Only the parts that are needed, so the cost per room stays low
- Built exactly for the system, with no unnecessary features
- Industrial-grade wired connection, independent of Wi-Fi
- Each unit is tied to a room number, so installation and repair are simple
- Supply, quality and support stay fully under our control
Proven in testing, not just on paper
Measured on one computer with a single GPU and a Dahua IP camera.
- Real-time speed
- 21 to 26 frames per second; a full frame is processed in about 25 to 29 ms
- Long-run stability
- 55,852 frames through a 43-minute lesson, and the stream never dropped once
- Headroom
- The GPU is loaded to about one third on average
- Automated tests
- All 65 logic tests pass
- Crash recovery
- If the program stops suddenly, the state of the lesson is restored
Example: a full system for 360 rooms
131 875 AZN
about 366 AZN per room on average
- Hardware
- 102 475 AZN
- Installation
- 14 400 AZN
- Software, one-time
- 15 000 AZN