Byteception

Tea Estates · Workforce · Fraud Detection · Computer Vision

Attendance Audit

Catching proxy attendance across tea estate workforces by checking every attendance photo, and naming whose face it really is.

Client
Confidential
Industry
Tea plantations
Our role
Engineering
Use case
Attendance fraud detection
FIG.01 · audit reportillustrative, anonymised
WorkerAttendance photosStatus
Worker A
100.0%
99.9%
99.8%
Pass
Worker B
100.0%
12.4%
99.6%
Fail
Worker C
100.0%
99.7%
Pass
Worker D
100.0%
99.9%
99.9%
99.5%
Pass
Worker E
100.0%
No face
Fail

The problem

Attendance you can't take at face value

Tea estates run on large workforces, and workers' pay is tied to the days they are marked present. Attendance is recorded with a photo, which should settle who showed up. In practice, one worker can be marked present on someone else's behalf, and the photo in the register is quietly of the wrong person.

Spotting that meant going through attendance registers by eye, worker by worker, photo by photo. It is slow and tiring, and a proxy is easy to miss on page after page of near-identical rows.

The approach

Let the machine look, let people decide

  1. Read the register

    Each worker's attendance photos are paired with their record automatically, so nobody has to sort images by hand.

  2. Compare every face

    Each attendance photo is checked against the worker's reference photo and given a similarity score. Photos with no clear face are flagged too.

  3. Report the exceptions

    Every worker gets a clear pass or fail, and every mismatched face is traced back to the worker it actually belongs to.

FIG.02 · cross-matchillustrative, anonymised

Photo in Worker B's attendance

Did not match

Identified as Worker K

99.9% similarity

The outcome

A clear audit trail for every worker

Each estate's register now produces a readable report: every worker, every attendance photo, a similarity score and a clear status. Instead of hunting for proxies, auditors start with a short list of suspects, and know who stood in for whom.

One run, not a manual review

An estate's whole attendance register is audited in one batch run instead of a photo-by-photo review.

Review only exceptions

Auditors skip the workers who pass and focus on the few cases that need attention.

Answers, not just flags

A suspected proxy comes with the name of the worker whose face it really is.

Have a product to build?

Talk to us

Fully remote, based in Bengaluru, India

© 2026 Byteception Software LLP. All Rights Reserved.