Counts you can act on
Footfall, dwell, queue length and conversion by hour and by entrance, reconciled against till or gate data where it exists.
AI Core — Where we lead
Computer Vision & Analytics
The cameras you already own, turned into counts, alerts and a number your operations meeting can use.
- Footfall, dwell and heatmap analytics
- ANPR and face recognition pipelines
- Edge inference and NVR integration
- Real-time alerting dashboards
- Engagement shape
- Site audit and a single-site pilot, then a rollout priced per site.
- Typical duration
- Three to six weeks for the pilot site, faster once the pattern is proven.
- Starting point
- One site, its camera list, and the question you want answered every day.
Overview
What the work involves.
Most sites do not need new hardware. They need inference running close to the cameras already on the wall, and a pipeline that survives bad light, monsoon glare and a lens nobody has cleaned since installation.
We run models at the edge so video never leaves the site, and send only events and counts upstream. Bandwidth stays low, footage stays inside your boundary, and a broken internet link delays a report rather than losing a day of data.
Accuracy is quoted per site after a calibration period, never from a datasheet. A model that reads number plates well in a lab still has to meet plate variation, night glare and a lorry parked at an angle, so the figure we publish is the one measured on your footage.
Outcomes
What you get.
Alerts within seconds
A safety breach, an occupied restricted zone or an unrecognised plate raises an alert at the site itself, not after an upload completes.
Footage that stays put
Inference runs on site hardware. Only events, counts and thumbnails travel, which holds both bandwidth and privacy exposure down.
A per-site accuracy figure
Measured on your own footage during calibration, with the failure cases written down rather than averaged away.
Deliverables
What lands in your hands.
- Camera and placement audit with coverage gaps listed
- Edge inference deployment on site hardware
- Detection and tracking models tuned on your footage
- ANPR or recognition pipeline where the use case needs one
- Integration with your existing NVR and CCTV
- Analytics dashboard with hourly and daily reporting
- Alerting rules routed to the people who act on them
Approach
How we run it.
Audit the cameras
Angle, lighting, resolution and frame rate for every view in scope. We mark which cameras can answer the question and which need moving before anything is trained.
Calibrate on your footage
Models are tuned and thresholds set on samples from your own site, across the shifts, seasons and weather that actually cause the errors.
Deploy to the edge
Inference on site hardware, events buffered locally and forwarded upstream. A dropped link queues the data instead of discarding it.
Reconcile, then report
Counts are compared against till, gate or manual records for a fortnight. The dashboard goes live once the two agree, not before.
Each phase ends with something you can read and act on. If the evidence says stop, stopping there is a supported outcome rather than an awkward conversation.
Stack
What we build it with.
- YOLO
- PyTorch
- NVIDIA Jetson
- DeepStream
- OpenCV
- ONNX Runtime
- RTSP
- TimescaleDB
Tooling is a decision we make per project, against your constraints and whatever your team already runs. Nothing on this list is a requirement, and we will work inside your existing stack where it holds up.
Questions
What clients ask about Computer Vision & Analytics.
Usually not. Most installed CCTV carries enough resolution for counting and tracking. The audit names the specific views that need a new angle or lens, and that list is normally short.
On your existing NVR. Our pipeline reads the stream, runs inference on site, and sends only events, counts and thumbnails onward. Raw footage stays on the premises unless you ask otherwise.
It depends on the lane, the vehicle speed and the light, which is why we quote per site after calibration. Plate format variation and night glare are the two cases we tune for first.
Ready to scope Computer Vision & Analytics?
Tell us where the work sits today and what is holding it up. We will come back with the shape of a first phase, what it would prove, and what running it takes.
+91 97915 97993