Face Recognition Attendance
Contactless attendance that cannot be proxied, with live in/out records, late marking and payroll-ready exports.
Result: Buddy punching ends and attendance disputes stop.
08 · AI Vision Systems
You already have the cameras. They just aren't watching.
Most CCTV in India is a recording device you look at after something has gone wrong. Vision AI turns it into a system that tells you while it is happening — and does the same for attendance, quality control and access.
- Starting at
- ₹1,50,000
- Typical timeline
- 6–12 weeks
- Services in this line
- 7
The problem
Footage nobody watches has almost no value. The value is in real-time alerts and in automating the counting, checking and identifying that people currently do by eye.
What we build
7 services in AI Vision Systems
Each one is a page of its own — pick the closest match, or tell us your problem and we'll point you at the right one.
CCTV Monitoring Agent
Watches your existing feeds continuously and alerts on what you define — intrusion, loitering, missing safety gear, crowding, a fallen person.
Result: You find out during the incident, not the next morning.
Number Plate Recognition (ANPR)
Automatic vehicle entry and exit logging for gates, parking and yards — tuned for Indian plate formats and conditions.
Result: Barriers that open for residents and log everyone else.
Quality Inspection System
Inspects every unit on the line for defects, dimensions or missing components at a speed and consistency no human can match.
Result: Defects caught at the station, not at the customer.
Product Image Tagging & Cataloguing
Auto-generates attributes, categories, alt text and descriptions from product photographs across a whole catalogue.
Result: A catalogue that is actually searchable, built in days.
Virtual Try-On
Customers see eyewear, jewellery, apparel or cosmetics on themselves through their camera before buying.
Result: Higher confidence to buy and materially fewer returns.
Photo Damage Assessment
Estimates damage severity and likely repair cost from photographs, with a consistent standard across every assessor.
Result: Claims processed in minutes with far less argument.
What you get
Included in every engagement
- Model trained or tuned on your actual environment and lighting
- Real-time alerting to WhatsApp, app or dashboard
- Works with your existing IP cameras where possible
- Edge deployment option so video never leaves your premises
- Web dashboard with search and history
- Accuracy benchmark report before go-live
How it's built
Typical stack
- YOLO / detection models
- OpenCV
- ONNX / TensorRT
- NVIDIA Jetson or cloud GPU
- RTSP camera feeds
We pick tools per project rather than forcing every client onto the same stack. If something in your business already works, we build around it instead of replacing it.
In depth
AI Vision Systems: what you should know before you commit
Most CCTV in India is a recording device that gets reviewed after something has gone wrong. Vision AI turns it into a system that tells you while it is happening — and brings real obligations under the DPDP Act.
Footage nobody watches has almost no value
The typical Indian business has more cameras than it had five years ago and reviews the footage less. Recording is retrospective by nature: it helps you establish what happened, long after you could have done anything about it.
The shift with vision AI is from recording to alerting. The same cameras, watched continuously by software that understands what it is seeing, generate a notification at the moment an event occurs — an intrusion, a person on the floor, missing safety equipment, a crowd forming where one should not.
The second, quieter value is automation of counting, checking and identifying — work people currently do by eye, slowly and inconsistently.
The applications that actually get deployed
| System | Where it fits | What changes |
|---|---|---|
| Face recognition attendance | Factories, offices, schools, sites | Proxy attendance stops; payroll disputes end |
| CCTV monitoring agent | Warehouses, societies, retail | You find out during the incident, not next morning |
| Number plate recognition | Societies, malls, industrial estates | Gates open for residents, log everyone else |
| Quality inspection | Manufacturing and packaging | Defects caught at the station, not at the customer |
| Product image tagging | E-commerce with large catalogues | A catalogue that is actually searchable |
| Virtual try-on | Eyewear, jewellery, apparel, beauty | Higher confidence to buy, fewer returns |
| Photo damage assessment | Insurance, rental, logistics | Claims in minutes, with a consistent standard |
Indian conditions are the actual engineering problem
Vision models that perform well on benchmark datasets often struggle on real Indian deployments, and the reasons are mundane rather than exotic.
- Lighting. Mixed daylight and tube light, heavy shadows, glare off polished floors, and sites that switch to generator power with different colour temperature.
- Camera placement. Cameras installed for retrospective review are often mounted too high, too wide, or angled for coverage rather than recognition.
- Number plate variety. Indian plates vary in font, spacing, mounting and condition far more than the datasets most ANPR models are trained on.
- Dust and weather. Lens contamination degrades accuracy gradually, which is worse than failing outright because nobody notices.
This is why we do a site survey before quoting, and why we benchmark accuracy on your site before go-live rather than quoting a number from a datasheet. If a camera needs replacing or repositioning, you should hear that before the project starts, not after.
Your footage does not have to leave the building
For attendance and surveillance, we default to edge deployment — the models run on hardware installed on your premises and no video goes to the cloud at all. Only events and metadata are transmitted.
This costs a little more in hardware and removes an entire category of risk and objection. For most Indian deployments it is the right choice, and for anything involving employee biometrics we consider it close to mandatory.
Will it work with our existing cameras?
Usually. Any camera exposing an RTSP stream can be used, which covers the overwhelming majority of IP CCTV installed in the last decade. Analogue systems with a DVR can often be tapped as well.
The cameras that need replacing are the very low-resolution ones and the badly positioned ones — and we will tell you which during the site survey, with a specific list, rather than discovering it halfway through the build.
Delivered with every vision deployment
A model trained or tuned on your actual environment, real-time alerting to WhatsApp or a dashboard, integration with your existing cameras where possible, an edge deployment option, a web dashboard with search and history, and an honest accuracy benchmark taken on your site before go-live.
FAQ
Questions people actually ask
Will it work with our existing cameras?
Usually yes — any camera exposing an RTSP stream will work. Very low-resolution or badly positioned cameras may need replacing, and we will tell you that during the site survey rather than after.
Does our video footage leave our premises?
Only if you want it to. We can deploy entirely on-premise on edge hardware so nothing goes to the cloud — which is the right choice for most attendance and surveillance deployments.
Is face recognition legal in India?
It is used widely for employee attendance with consent. Under the DPDP Act biometric data carries specific obligations, so we build consent capture and retention limits into the system by default.
Other AI service lines
Next step
Let's work out what's worth building.
A 30-minute call. We'll tell you what we'd do, roughly what it costs, and whether it's worth doing at all.