AI / COMPUTER VISION
Facial recognition attendance, without cards or queues
A facial recognition attendance system marks people present by recognising their face as they arrive, instead of asking them to tap a card, punch a code or sign a register. Each person is enrolled once, which converts their face into a mathematical template rather than a stored photograph. At the entrance, a camera matches the face against those templates in under a second and writes an attendance record automatically. M3Stack builds these systems end to end — enrolment, real-time recognition at the door, liveness detection to stop photo spoofing, integration into the HR and payroll systems already in use, and the consent and retention controls that biometric data requires. Our product FaceMark AI recognises people the moment they walk through the door and marks them present with no badges to forget, no lines at the entrance and no registers to maintain.
01 / SCOPE
What the system actually involves
Enrolment
Capturing each person once and converting their face to a template. Done badly this is the source of every later false rejection, so it is worth getting right at the start.
Recognition at the door
Detecting and matching a face at walking pace, so a queue never forms. Camera placement and lighting matter as much as the model does.
Liveness detection
Distinguishing a real face from a photo, a screen or a mask. Without it, a printed picture held to the camera is enough to mark someone present.
HR and payroll integration
Attendance is only useful when it reaches payroll, leave and rostering. We build that path in rather than leaving it as a manual export somebody forgets.
Reporting
Ready-made reports for the questions that actually get asked — who was late, who is absent, overtime by team, attendance across sites.
Consent and retention
What is stored, where it lives, how long it is kept and how someone is removed. Biometric data carries obligations that a card reader never did.
02 / PROOF
FaceMark AI
Attendance without cards, codes or queues
FaceMark AI recognises people the moment they walk through the door and marks them present automatically — no badges to forget, no lines at the entrance, no registers to maintain. Accurate records and ready-made reports, with zero effort from staff.
It is built on the same computer-vision work behind our video platforms: real-time inference, video analytics and facial recognition running at the speed an entrance actually moves.
03 / THE PART BUYERS ASK ABOUT
Biometric data is not the same as a swipe card
The technology question — can it recognise faces accurately — is usually the easy one. The question that decides whether a deployment survives contact with an HR director, a works council or a legal team is what happens to the data.
These are the decisions worth making deliberately, before the system is built rather than after someone asks:
- Templates, not photographs. A face can be stored as a mathematical template from which the original image cannot be reconstructed. That is a meaningfully different risk profile from a folder of employee photos, and it is worth being explicit about which one a system holds.
- Where the data lives. On-premise deployment keeps biometric data inside the organisation's own network. Cloud is simpler to run. This is a genuine trade-off and the right answer differs by organisation.
- Consent and the alternative. People need to know what is being collected, and there is usually a requirement to offer a non-biometric option for anyone who declines.
- Retention and deletion. How long templates are kept, and what happens when someone leaves. A system with no deletion path accumulates biometric data on former staff indefinitely.
- Access and audit. Who can see attendance data, and whether access to it is itself logged.
We build these as configurable decisions rather than baking one answer in, because the correct setting depends on the organisation, the jurisdiction and the workforce. What the specific obligations are in your case is a question for your own legal advisers — our job is to make sure the system can actually implement whatever answer they give.
04 / HOW IT RUNS
Four stages, working software every week
Discover
Headcount, entry points, shift patterns, the HR system already in use, and what the organisation is obliged to do with biometric data.
Architect
Camera placement, on-premise or cloud, confidence thresholds, and how attendance reaches payroll.
Build
Short cycles with working software every week, usually piloted at one entrance before rolling out.
Launch & support
Enrolment of the full workforce, then production support — because an attendance system that fails at 9am fails loudly.
05 / QUESTIONS
Common questions
How does a facial recognition attendance system work?
Each person is enrolled once, which converts their face into a mathematical template rather than a stored photograph. At the entry point a camera detects a face, converts it the same way, and compares it against the enrolled templates. A match above the confidence threshold writes an attendance record with a timestamp. The whole exchange takes under a second and needs no card, code or physical contact.
Is it more accurate than a fingerprint scanner?
In practice it is more reliable in the conditions where fingerprint readers struggle: wet, dirty, worn or cut fingers, gloved hands, and high-traffic entrances where queuing at a single sensor is the bottleneck. Face recognition is contactless and works at walking pace, so throughput at a busy entrance is far higher. Accuracy depends on camera placement, lighting and the confidence threshold chosen for the deployment.
Can it be spoofed with a photograph?
Not if liveness detection is built in. Liveness checks distinguish a real face from a photo, a screen or a mask, using signals such as depth, texture and micro-movement. Any attendance system tied to payroll needs liveness, because without it a printed photo held to the camera is enough to mark someone present.
Does it store photographs of our employees?
It does not have to, and generally should not. A well-designed system stores a mathematical template derived from the face rather than the image itself, so the original photograph cannot be reconstructed from what is held. Storage location, retention period and whether raw frames are kept at all are deployment decisions that should be made deliberately, in line with your obligations for biometric data.
Can it integrate with our existing HR or payroll system?
Yes. Attendance data is only useful once it reaches payroll, leave management and rostering. We build the integration layer as part of the system rather than leaving it as a manual export, so shifts, overtime and absence flow into the systems already in use. Integration method depends on what your HR platform exposes.
06 / ALSO FROM M3STACK
Related work
07 / OPEN CHANNEL
Tired of queues
at the front door?
Tell us your headcount, your entry points and what HR system you run. We'll reply within 24 hours.
team@m3stack.com