Attendance and time-on-task
Smart Attendance Tracker
Measured98.7%
Built on face recognition: arrival time, lateness and absence are registered and totalled automatically. It takes the register back off the teacher entirely.
Mergen AIFar-sighted education.
Is the camera in your school only making a recording?With Mergen AI it becomes something that works for every student in the room — one that is safer to sit in, and better to learn in.
Schematic — not captured footage
I / V
Why would a school need Mergen AI?
Our purpose
Safeguarding the next generation, intelligently
Academic analysis
AI analysis of where a student is falling behind and where their attention goes, used to build the habit of studying independently rather than to rank them.
Emotion analysis AI
Real-time analytics on students' emotional state and behaviour, used to get ahead of bullying and keep the room a safe place to be.
Student evaluation report
A complete assessment of each student, compiled automatically — it gives teachers their time back and puts a real basis under the decisions they make.
The problem
Mergen AI sees the hidden risk and the quiet falling-behind that no one in the room can, and proposes something to do about it
This is what the system is pointed at. A child who is quietly excluded, a run of absences nobody joined up, a class drifting from the lesson — none of it announces itself, and all of it is visible in patterns that build up over weeks.
85%of students have their right to learn and to be safe in school compromised in some way
II / V
What the system does
Mergen detector
Six things it watches for
One camera, one PC in the building, six detectors sharing a single frame budget. Attendance is the one with a measured figure behind it; the rest are described by what they do, not by an accuracy number nobody has taken yet.
Safe environment and bullying detection
Safety & Anti-Bullying Monitor
Analytics on body movement and micro-movement, used to flag the risk of bullying, of violence, and of physical confrontation early enough to act on.
Engagement and participation analytics
Classroom Engagement Tracker
Shows where attention holds and where it goes across a lesson, along with who is being left behind — the raw material for making the teaching better rather than for grading the class.
Exam integrity
Exam Integrity Guard
Watches for the gaze and the movements that go with copying, and for phones, so an exam is scored on what was actually done in it.
Uniform and dress code analytics
Uniform & Dress Code Analytics
Checks uniform and dress standards against the school's own rule, and reports the exceptions rather than every student.
Well-being and stress detection
Student Well-being Analytics
Analyses changes in a student's expression over time, so prolonged isolation, low mood or sustained stress reaches the school's counsellor early rather than late.
Six detectors, six letters of the name — each letter is a mark, not a claim about the row it sits on
One lesson, end to end
The path a single lesson takes, from the camera to a mark a parent can read.
- 01Capture
- 02Detect
- 03Match
- 04Record
- 05Teacher corrects
- 06Correction persists
The last two steps are what make the loop trustworthy rather than merely automatic: a teacher can overrule the system, and it is the correction that persists.
In the school
Where the video stops
One PC per school, not one per classroom. Cameras reach it over the local network and raw frames go no further; only metadata leaves the building.
- 01In the school
- 02Local video
- 03Metadata only
- 04Cloud
- 05Corrections back
Corrections travel the other way. A teacher edits in the cloud and the box picks the change up on its next sync.
III / V
The same argument, at school scale
Nobody signs themselves up
Every account is created one level above and handed over. There is no self-registration anywhere in the product.
IV / V
Taking it apart in public
What is already true
- 98.7%
- of faces identified from the ceiling angle
- 27
- automated checks passing end to end
- 0
- frames of classroom video that leave the building
What acquisition looks like when everything is in your favour
match
Head-on, evenly lit, one subject filling the frame: the pattern lands, six landmarks settle, and a 512-dimensional embedding comes out clean. This is the demo every vendor shows you, and it is the easy case.
Mergen never gets this view. The camera is mounted high above the board and looks down the length of the room — heads are tilted to desks, turned to neighbours, or behind the row in front. Getting to 98.7% from that angle is the engineering, and it is why the number is quoted for the ceiling mount rather than for a head-on gate that nobody installs in a classroom.
One pass of the pipeline
Sixteen figures are detected in a single loop. Fifteen resolve by face; the remaining one does not, and is carried by seat occupancy instead. The similarity reading beside each box is illustrative — a comparison against one gallery entry, not a measured rate.
present16 / 1615 by face
- presence
- 16 confirmed
- identity
- 15 matched
- emotion
- not inferred
- attention
- not measurable
- stress
- not scored
- person det.
- face recognition
- seat occupancy
- exam mode
- incident scan
- attention
The faces above are drawn differently on purpose — bored, delighted, furious. That difference is now something the system reads rather than something it ignores: the well-being detector tracks how a student's expression changes across weeks, and reports a sustained shift to the school's counsellor rather than a snapshot to a parent. What it never does is publish a mood as a fact about a child on any given day. A single frame is not evidence of anything.
One lesson, thirty-five seats
One lesson, thirty-five seats
- Identified by face27
- Present — seat only1
- Empty7
Schematic — not captured footageOf the 28 occupied seats, 27 are identified by face. The one that is not is carried by seat occupancy, which is what keeps the register complete on the frames where a face never becomes usable.
V / V
What it ends up with
What a parent opens
Three students, three weeks of attendance. No grades, no ranking, no score — the system holds none of those. It counts lessons, and counting turns out to be the half of the promise the evidence supports.
Чирж эргүүлнэ · ← →
9б анги
Ирсэн08:04Өнөөдөр · хаалганд
76%товлогдсон хичээлийн хэдэн хувьд ирсэн
3Шалтгаангүй
1Хоцорсон
84Хичээл
2Дараалан
Хичээлийн ирцөдөр тутамшалтгаангүй
04.2004.2905.11
дундаж 76%хамгийн бага 33%бүтэн 5/15хандлага −13%
98.7% нүүрийг таазны өнцгөөс таньсан хувь0 хичээлийн видеоны фрейм барилгаас гардаг
This is attendance, not academic performance. There are no grades anywhere in the system and nothing here scores a child. What it can say is how many lessons went missing in three weeks and how many of those nobody accounted for — which is a sentence a parent can act on tonight, and which no model had to guess at.
Get in touch
If you are hiring, evaluating this for a school, or you think something on this page is wrong, write to me. Messages go to one inbox and are answered by a person.
This form sends what you type in it and nothing else. No analytics on the submit, no third-party embed, no cookie set, and your address is used to reply to you and for nothing further. The same standard the rest of this page argues for applies to the one form on it.