Apar Drishti — Vision AI
Apar Drishti is Aparsoft Private Limited's vision AI for the cameras a site already has: CCTV, drones and radar. It runs on the customer's own servers, with no cloud and no new hardware. Drishti is 12+ working prototypes, built to your site. Each project starts from the prototype closest to the need and is configured to that site's cameras, zones and rules.
Cameras everywhere. Drishti watches every feed. Nobody can watch every screen all night, or sit through last week's recordings. Drishti does both: it alerts your team as it happens, and turns hours of footage into counts, timelines and answers.
12+ working prototypes, built to your site · Your cameras, the ones you have, no new hardware · Your servers, no cloud, footage never leaves your premises
It Sees. It Explains. It Alerts.
- It sees: fire and smoke, people, vehicles, helmets and vests, on every feed at once
- It explains: a note in plain words saying what it saw, how serious it is and what to do
- It alerts: on screen, on WhatsApp, to your own system, or as a siren on site
On the Ground (CCTV)
| Prototype | What it does |
|---|---|
| CrowdIQ | Counts people and alerts at your limit. Each camera has its own limit, and every alert is kept with its pictures, the AI's note and a CSV export |
| FireIQ | Fire and smoke on every feed: seen, explained, alerted |
| IDS | Draw a zone on a camera. Restricted, loiter and gate zones; counts in and out at a gate |
| Site safety | Helmet and vest on every worker, by your site's own rules; each miss logged with a picture |
| Person Tracker | One ID per person across your cameras, and the path they took. Matching is by appearance: no faces, no biometrics |
| Face Access | Entry without a card, for people who agree to it: enrol once, verified at the door, every entry logged |
Answers From Footage, Not Only Alerts
- ShiftIQ answers who was where, and for how long. Every worker gets a number that follows them across cameras. Zones are marked once on a camera picture. You get a timeline and timesheet per worker as a CSV. It uses numbers, not names, with heads blurred, and lists every identity decision.
- StoreIQ turns store cameras into answers. It counts shoppers in and out, shows a heatmap of where they walk and stop, flags a queue at the counter, and keeps staff out of the count. No face recognition.
Sentinel Sky: Drone and Radar
- Sentinel Sky · Drone puts any drone on the map. It reads the position newer drones send inside the video, and the numbers older drones burn onto the picture. Pick any pixel to get its place on the map, checked against the drone's own record. Working prototype, public test footage.
- Sentinel Sky · Radar sees through cloud and dark. It finds ships, aircraft, vehicles, bridges and harbours. It was trained on 116,000 images from ten public radar collections, checked on images it never saw, and runs on one GPU, fully on-premises, with no internet. Working prototype, public benchmark imagery.
YOLO Vision Studio: Our Free App, in the Open
Open source, 400+ GitHub stars. It is built on Ultralytics YOLO. Every object gets an ID and keeps it, it counts after five frames, and it never knows who anyone is. Code: github.com/aparsoft/yolo-streamlit-detection-tracking.
The First Questions a Site Head Asks
| Question | Answer |
|---|---|
| Do we need new cameras? | No. It works with the cameras you already have (RTSP or USB). |
| Does our footage go to the cloud? | No. Everything runs on your own servers. |
| Is it a finished product? | It is 12+ working prototypes. We build the one you need to your site. |
| Does it know who people are? | No. Only Face Access does, for entry that people agree to. |
How a Project Starts
- Bring us a clip from one of your cameras, your drone's video or a scene from your sensor
- See it work on your own footage
- Built to your site: your cameras, zones and rules, on 12+ working prototypes
- Live on-prem: on your servers, alerts go to your team
No price list: every Drishti project is built to the site and quoted per project. [email protected] · +91 89040 64878 · aparsoft.com/products/apar-drishti.
House rules for anything printed about Drishti:
- Say "working prototypes, built to your site". Never say "deployed at", "live product" or name customers.
- Print no accuracy, speed or "real-time" figures.
- Face recognition appears only as Face Access.
- Radar pictures are labelled "public benchmark imagery".
- Film scenes of the 3D site are labelled "Illustration"; app screenshots are labelled "Real screen".
Apar Drishti runs on the same reusable Apar AI Engine (vision AI on-premise, agentic workflows, multi-tenant deployment) that powers all of Aparsoft's products. Aparsoft Private Limited is a DPIIT-recognised, Bengaluru-based AI startup. Founder: IAF Veteran · AI research at CAIR-DRDO.