Performance Tracker turns the cameras you already have into a clear, fair picture of every working day — who is in, where they are, what they're doing, how customers are served and what happens at the till. For offices, restaurants, cafés and shops. Every minute explained, nothing guessed silently.
The same cameras, read for what matters to you — the counter in a café, the entrance at the start of a shift, the desks in an office.
Cash-handling alerts, customers left waiting, possible dropped food and a plain daily summary for the owner.
Restaurant CCTV analytics → AttendanceArrivals and departures from your entrance cameras — no fingerprint machine, no queue, no buddy punching.
CCTV attendance system → OfficesTime at work, phone use, meetings and breaks — judged against each person's own normal, with nothing installed on their PCs.
Employee monitoring with CCTV →Four stages run on their own every hour. Each one is built to say what it saw — and to say so when it couldn't see enough.
Plugs into your existing NVR. Every camera is sampled through the day from its recordings — even ones nobody is watching — over a private encrypted tunnel.
Finds every person in frame, reads body posture, and recognises enrolled employees by face — then follows them across cameras and the seat they sit at.
Combines posture, hands on keyboard or phone, movement, company and the room to classify each moment — and marks whether it was directly seen or inferred.
Builds a personal profile of how each employee works and what their normal day looks like, and improves from every correction your team makes.
Most CCTV is only ever watched after something goes wrong. Performance Tracker works on the footage your NVR already records.
Arrivals and departures from the entrance cameras, then a minute-by-minute picture of every room each person spends time in.
Every sampled moment is classified from several signals together — body posture, hands on the keyboard or mouse, a phone in hand or at the ear, movement between frames, who else is nearby, and the room. When the evidence isn't there, it says Unclassified instead of guessing.
Hands on keyboard or mouse, or a posture that matches how this person works at their computer.
A phone gripped in the hands or held to the ear — counted separately, even at the desk. A phone lying on the desk doesn't count.
With others away from their own desk, or turned towards someone. Neighbours in an open-plan office don't trigger it.
Away from their own usual working posture with hands off the desk — leaning back, arms crossed, hands behind head.
Seen in the courtyard, parking or outside the work areas, rather than just walking through.
Standing and moving between places — corridors, lifts, lobbies.
Body hidden or no clear sign either way. Never counted as work or as idling.
Directly visible in the snapshot: a phone in hand, hands on the keyboard, or set by a reviewer. Shown as a solid bar.
The most likely reading from posture and context — a judgement, always shown striped so nobody mistakes it for an observation.
In a restaurant or café the cameras over the counter see more than any manager can. Performance Tracker turns them into a short list worth checking at the end of the day.
Each employee's day as stretches of activity with times, rooms, posture and how sure the system is — with the photo behind every sample one click away.
Recognises enrolled staff from a photo, then learns their look on your own cameras — more angles, better recognition every week.
Joins one person's sightings across rooms by face, clothing and the seat they're at — so a back turned to the camera still counts.
Learns how each person works — someone who works leaning back isn't flagged for it; someone who sits upright is noticed when they don't.
Knows each person's usual hours, places and breaks, scores every day against it and flags unusual hours — arriving late, a missing afternoon.
Ten one-click questions a day on the samples the AI was least sure about. Every answer corrects the report and teaches the model.
Writes each employee's day up in plain language — seat and work, movement, phone and breaks, what was unusual and what needs attention.
Watch any camera's recording in the browser at up to 16× speed, with audio up to 4×, protected by its own login.
Time and activity per office, meeting room and common area — for each person and for the whole team.
An hourly health check of cameras, database, detection quality and learning progress — and it restarts anything that's down.
Analytics about people only work if they're honest. These principles are built into how every number is produced.
Confirmed and AI-inferred time are always separated, and "Unclassified" is a first-class answer — never quietly turned into work or idleness.
Posture and routine are compared with each person's own history, not one rigid rule for everyone.
Every reading can be corrected by a reviewer, and the correction overrides the AI and teaches it.
It sees posture and objects from the cameras you already have. It never records conversations or what's on anyone's screen.
Runs on a PC in your office or a private cloud server you control. Footage never leaves your infrastructure.
HTTPS, a sign-in for every page, and the NVR reachable only over a private encrypted network — never on the public internet.
Runs on a standard server — no GPU, no new cameras. Choose where it lives. Serving offices, restaurants, cafés and shops in the UAE, Pakistan and the GCC.
No. It works with the recordings of Hikvision and Dahua NVRs over standard RTSP, and runs on an ordinary server — a 4-core PC is enough; no graphics card is needed.
Yes. Besides staff attendance it watches the till and the counter — a hand going from the till to a pocket, the till used with nobody to serve, customers waiting with nobody coming to the till, possible dropped food — and turns it into a plain daily summary for the owner.
Yes. You enrol each employee with a few photos; the entrance cameras mark arrivals, and every later sighting is linked to the same person by face, clothing and where they work. Anyone the cameras miss is listed for a one-click check.
Directly visible evidence — a phone in hand, hands on the keyboard — is marked as confirmed. Everything inferred from posture is labelled as such, and the system answers "Unclassified" when a snapshot doesn't show enough. Accuracy improves as your reviewers answer the daily review list, and the dashboard shows the learning progress honestly, including where it isn't better than chance yet.
No. It only uses the video your cameras already record, and only reads posture, position and visible objects. "In conversation" means people are together, not that anything was heard.
You enrol each employee with a clear photo. It matches faces on camera, learns more of their camera look from confirmed matches, and links face-less sightings by clothing and the seat they're at. Reviewers confirm or correct identities in one click.
Every hour, each camera is sampled every ten minutes from its recordings and the reports are rebuilt. Entrance cameras mark arrivals in real time, and till and counter alerts appear as they happen.
On the server you choose — a PC in your office or your own private cloud server. Access is over HTTPS with a sign-in, and the NVR is only reachable over a private encrypted network.
See it running on your own cameras — tell us a little about your team and we'll walk you through a live dashboard.
Tell us a little about your workplace and we'll get back to you to arrange a live walkthrough.
We'll be in touch shortly to schedule your demo.
