dayFlow

Face recognition attendance

Know that the person clocking in is the person on the payroll

Every clock-in comes with a selfie that is matched against the face enrolled for that person, so clocking in for a colleague stops working. Doubtful matches are not guessed at: they go to a manager with the photo beside them.

  • A selfie matched to the worker's own face
  • Photos of photos and copies are caught
  • Doubtful matches wait for a manager, with the evidence
Under 1 minute
to register a face
1 : 1
matched to the person's own face
3 safeguards
liveness, replay and photo-quality checks

In plain words

What is face recognition attendance?

It is a way of recording that a particular person was present, not just that a phone or a card was. When someone clocks in, the app takes a selfie and the server compares it with the face that was registered for that person. A match is recorded as a normal punch; a doubtful result is recorded too, but set aside for a manager to look at.

It matters most where hours are paid and nobody supervises the gate. A shared card or a phone passed to a friend proves nothing, while a face is hard to lend. dayFlow compares one face with one registered person (not a search through everyone), and adds checks that catch a photo of a photo, a screen held up to the camera, and a copy of the enrollment picture.

What you get

Face recognition attendance, feature by feature

Registered once, checked every time

A manager registers each person with their consent. After that every clock-in compares the new selfie with those registered templates, and the result is saved with the punch: matched, low confidence, mismatch, or not enough of a face to judge.

In practice

  • A short camera clip or three to eight photos at registration
  • Consent is recorded, and can be withdrawn
  • The match result and score are kept on the punch

Fraud that photos usually allow, caught

A face recogniser that only compares faces can be fooled by a photograph. dayFlow also looks for signs that the camera saw a picture or a screen, and for a selfie that is a pixel-for-pixel copy of an enrollment photo, which a live person never produces.

In practice

  • Live-person (anti-spoof) check on every selfie
  • Replay detection for copied pictures
  • Blurry, tiny or faceless photos are marked unusable, not guessed

Doubtful is not the same as guilty

A bad angle or poor light should not cost anyone a day's pay. Anything short of a clear match is kept, flagged, and put in one review queue with the new selfie beside the registered face, so a manager can decide in seconds.

In practice

  • Accept or reject from the review queue
  • A rejected punch stops counting towards hours
  • The decision and the person who made it are recorded

Private by design

Face data is sensitive, so it is treated that way. Templates are stored encrypted, only doubtful selfies are kept for review and they are deleted after a limited time, and a person's face data can be erased on request.

In practice

  • Encrypted templates, not photos on a disk
  • Review photos purged automatically
  • Erase one person's face data whenever asked

The difference

What changes when you use it

Without dayFlow

With dayFlow

A friend clocks in for you and nobody notices

The selfie must match the face registered for that person

A printed photo or a phone screen is held up to the camera

Liveness and replay checks flag it for review

Disputes are settled by memory

Every doubtful punch keeps its photo, score and decision

Examples

How teams use it

  • Example 1

    A warehouse gate at shift change

    Forty people arrive within ten minutes. Each clocks in with a selfie in a second or two, and only the handful of doubtful ones reach the supervisor.

  • Example 2

    Contract cleaners at a client site

    The agency's workers are registered when they start. The client's hours are backed by face-verified punches, and a worker who was swapped for someone else shows up as a mismatch.

  • Example 3

    A manager approving on the dashboard

    Before signing off a week, the manager opens the exceptions list, sees the three flagged punches with their photos, and clears them before approving the days.

The details

What you can set

Registration
Short camera clip, or 3 to 8 photos, with consent
Matching
One face against the registered person (1:1)
Possible results
Matched, low confidence, mismatch, unusable, not registered
Spoof checks
Live-person detection and replay detection
Review photos kept
Only doubtful selfies, for a limited time, then purged
Face data
Encrypted at rest, erasable per person

How it works

Up and running in three steps

  1. 1

    A manager registers the employee's face, with their consent.

  2. 2

    The employee clocks in from the app and takes a selfie.

  3. 3

    The match happens at once. Clear matches are trusted; the rest wait in the review queue.

Who it is for

Made for teams like these

  • Companies with hourly or contract workers
  • Sites with no supervisor at the gate
  • Anyone who pays for hours and has met buddy punching
  • Teams that need proof for a client or an auditor

Part of a whole

How it fits with the rest of dayFlow

Face checks do the most work alongside location checks: a punch has to come from the right person and from the right place. The result feeds the review queue, the timesheet and, if you want it, a face check on the dashboard itself before sensitive approvals.

Questions

Things people ask

What happens when the face does not match?

The punch is kept, marked as a mismatch and sent to the review queue with the photo beside the registered one. A manager accepts or rejects it, and the timesheet follows.

Do we keep everyone's photos?

No. Face templates are stored encrypted. Only doubtful selfies are kept for review, for a limited time, and anyone's face data can be erased on request.

Is it hard to register people?

A manager records a short camera clip or takes three photos, with the employee's consent. It takes under a minute per person.

What if someone has not been registered yet?

Their punches are still recorded, with a note that they were not face-checked, so nobody is blocked from clocking in on their first day. A manager registers them when they can.

Can the same face be registered for two people?

No. If a face already belongs to another person, registration is refused, which stops one person holding two identities.

Does it work in poor light or with a mask?

A clear view of the face is needed. When the photo is too dark, blurry or small to judge, it is marked unusable and reviewed by a person instead of being failed automatically.

Can employees register themselves?

No. A manager or admin registers faces for employees, so a person cannot enrol someone else's picture as their own. Admins and managers can register their own face for the dashboard face check.

Try Face recognition attendance with your team

Set up your locations, register faces and let your team clock in. Your company gets its own address.