Apar Drishti

See threats before they become incidents.

Fire detection. Crowd safety. PPE compliance. Intrusion alerts. Face access. Drone video on the map, and radar imagery. Working prototypes you can see running, built to run on your own servers.

Apar Drishti: video AI on your own servers — One site, one night: smoke at a dock, a zone breach, a missing helmet. Seen, explained and alerted.

12+

Working CV systems

Simultaneous camera feeds

On-prem

No cloud dependency

Real-time

Detection and alerts

How it works

One platform. Deployed for your specific problem.

Every Drishti system — fire detection, crowd monitoring, PPE compliance, and more — runs on the same foundation. That means faster deployment, consistent performance, and the ability to expand to new use cases without starting over.

Works with your existing cameras — RTSP, USB, or live feeds. No hardware replacements required.

Alerts reach your team the way they already work: on screen, over the network, via webhook, or as physical outputs.

Runs entirely on your own servers. Your footage never leaves your premises.

When something happens, AI explains what it saw — the situation, the severity, and what to do next.

Sentinel

From the drone, and from orbit.

The same on-premises engine, pointed at two harder sources: drone video and radar satellite imagery. Watch each one work, on public footage and public benchmark imagery.

Sentinel · Sky

Any drone, on the map.

Working prototype · public test footage

A drone spots a vehicle. But a picture is not a position. Newer drones send their position as metadata inside the video, and we decode it. Older drones burn the numbers onto the picture, and we read them off the frame. Then pick any pixel: its ray meets the ground, and that point is a position on the map.

Reads both kinds of drone: metadata inside the video (MISB ST 0601 KLV), or numbers burned onto the picture.

Two independent readers: a fast one on every frame, and an on-premises AI that reads it again as a check.

The whole flight plays on the map in step with its video. Live streams, capture cards or files.

Checked against the drone's own record on a public test flight. Flat ground today; terrain models next.

Sentinel · Orbit

Through cloud and dark.

Working prototype · public benchmark imagery

Cloud and night blind a camera satellite. Radar (SAR) sees through both, but a radar picture is grey speckle, and a ship is a few bright pixels. Sentinel · Orbit finds ships, aircraft, vehicles, bridges and harbours in it.

Trained on the largest public radar benchmark: 116,000 images from ten radar collections.

Checked against the benchmark's own answer key, on images the model never saw.

Each object keeps its identity, frame after frame, as the sensor moves.

One GPU, fully on-premises, no internet. Today it reads image tiles; whole scenes with positions are next.

Sentinel · Orbit: through cloud and dark

The systems

12+ Vision AI systems. Each one solving a specific, real problem.

From fire safety to sports analytics, each Drishti system is built for a different operating environment. Same underlying platform. Different deployment. Ready when you are.

Fire & Smoke Detection

FireIQ

Your building is on fire. Who finds out first — your cameras, or the smoke alarm? FireIQ spots fire and smoke the moment it appears, judges how serious it is, and alerts your team before it escalates.

Detects fire and smoke across factories, warehouses, and facilities — day and night.

AI reads the scene and tells you what’s burning, how severe, and whether to evacuate — in plain language.

Alerts reach your team immediately: on screen, over the network, via WhatsApp, or through physical triggers.

FireIQ fire detection severity dashboard

FireIQ live detection view with severity counts and rapid incident visibility.

1 / 4

Crowd Safety

CrowdIQ

When a crowd is too dense for the space, seconds matter. CrowdIQ counts people in real-time, tracks density zone by zone, and flags a breach before it becomes a crisis.

Live headcount and density tracking across venues, campuses, and transport hubs.

Set capacity thresholds per zone — get alerted the moment an area becomes overloaded.

Gives safety teams proactive visibility so they can act before a situation escalates.

CrowdIQ — Bridge scene with crowd density analysis

CrowdIQ analyzing a bridge scene — person counting, density trends, and breach detection in live environments.

1 / 4

Intrusion Detection

IDS

Someone is in an area they shouldn’t be. How quickly do you find out? IDS watches your defined zones around the clock and raises an alert the moment a boundary is crossed — no human required.

Draw restricted zones directly on any camera feed. Alerts fire the moment anyone crosses them.

Detects loitering and suspicious dwell time — not just entry, but intent patterns.

Movement heatmaps show which areas see the most unauthorised activity over time.

IDS — Object detection for zone monitoring

Object detection monitoring a busy road — the same engine powers zone breach and intrusion alerts.

1 / 3

Cross-Camera Re-Identification

Person Tracker

After an incident, reviewing footage across multiple cameras manually can take hours. Person Tracker follows a person of interest across all your cameras automatically — delivering a full movement trail in minutes.

Tracks a single individual across 6 or more camera feeds simultaneously.

Maps their complete movement path for live tracking or post-event investigation.

Turns hours of manual footage review into a searchable, timestamped journey.

Person Tracker — Multi-camera feed on single dashboard

Multiple camera feeds processed on one dashboard — track individuals across cameras for journey mapping and post-event review.

1 / 3

Site Safety

Construction V2

One PPE violation can cost a life — and trigger a serious liability. Construction V2 watches every worker on site and flags anyone without the right safety gear, continuously and automatically.

Detects helmets, vests, and safety equipment across all active site viewpoints.

Tracks workers across multiple camera feeds without manual footage review.

Gives safety managers a live compliance signal instead of end-of-day guesswork.

Construction V2 — PPE kit detection and compliance

Real-time PPE compliance detection — helmets, vests, and safety gear verified against site requirements.

1 / 2

Access Control

Face Access

A fast, frictionless entry system for offices and checkpoints. Walk up, the system recognises your face, and access is granted or denied in under a second. No card. No PIN. No queue.

Three-stage recognition pipeline: detect, verify, decide. Completed in under 300ms.

Runs on standard hardware — no expensive GPU or specialist equipment required.

Suited for office entry, visitor management, and controlled access scenarios.

Representative access-control visual for front-desk, entry gate, and checkpoint scenarios.

1 / 2

More working prototypes

Sports analytics

FootballIQ

Player positions, possession flows and key moments from match video.

Multi-sport tracking

Sports Team Tracker

Players tracked across cameras, with teams told apart by jersey colour.

Shopper analytics

RetailIQ

How shoppers move through a store: where they slow down, and where they leave.

Operations

Workplace Insights

Depth and posture insights from a single camera.

Scene understanding

Segmentation Studio

Type what you want, such as “white cars”, and only those are outlined (SAM 3).

Open source

YOLO Vision Studio: our free app, in the open.

Built on Ultralytics YOLO. Find every object, outline each one, or read a body's pose. Or just type what you're looking for. In video, every object gets an ID and keeps it, and the counts are honest: an object counts only after five frames, and it never knows who anyone is.

Clone it, learn from it, build on it. When you need it on your own cameras, talk to us.

YOLO Vision Studio: our free, open-source app

Get started

Tell us the problem. We'll show you the system.

Whether you need to detect a fire, manage a crowd, enforce PPE compliance, or secure an access point — there's a working Drishti system for it. We start from running systems, not from a slide deck.