Your cameras already see everything. Arjuna makes them understand it.
Real-time video analytics that runs on the CCTV you have installed today. Detection, tracking, counting, zone alerts and safety-compliance monitoring on your own infrastructure.
The problem
A camera that only records is a camera that only helps you after the fact.
Most CCTV estates are passive. Footage is reviewed once something has already gone wrong, and reviewing it costs hours of somebody's day.
Nobody can watch forty screens.
Attention decays within minutes. Events that matter a vehicle in a pedestrian area, someone loitering by a restricted door, a worker skipping a safety step pass unseen in plain view.
The data is already there.
Footfall, dwell time, queue length, vehicle mix, occupancy, compliance rates. It's recorded every day and thrown away every month.
The market pull
$14.65B → $41.39B
Video analytics market, 2026 to 2031
Market research estimates
23.1% CAGR
Growth rate over that period
Market research estimates
Retrofit-first
AI on existing cameras, not rip-and-replace
Category shift
The category has moved. Until recently, video analytics meant buying specialised cameras with the intelligence baked into the hardware an expensive forklift upgrade. The current wave runs the intelligence in software against ordinary RTSP streams, which means a site can adopt it incrementally, one camera at a time. Enterprise vendors are building this into their VMS platforms; a cloud-native tier is selling the retrofit story directly. In India, the DPDP Act has made on-premise processing and face/plate redaction a procurement requirement. Arjuna sits in that retrofit tier, with on-premise deployment as the default.
How Arjuna works
Connect
Point Arjuna at your existing RTSP streams. No new cameras, no rewiring.
Detect
Every frame is analysed for people, vehicles, and objects, each with a tracked identity that persists as they move.
Apply your rules
Draw a zone, a line, or a restricted area once per camera. Arjuna watches it continuously.
Act
Live alerts, dashboards, and searchable history. Counts, dwell times, occupancy and compliance rates as numbers you can act on.
Use cases
Each clip is side-by-side raw feed on the left, Arjuna's analysis on the right. The contrast is the product.
Use case 01
People tracking & re-identification
Real CCTV from a fixed camera. 94 unique individuals tracked across 35 seconds, each holding an ID as they cross the frame and pass one another.
The foundation everything else is built on. Grainy footage, awkward angle, ordinary camera the same conditions as a real site.
Use case 02
Footfall & people counting
Live in-frame count alongside a cumulative unique count.
Two different numbers, and the gap between them is the product. "Eleven people right now" is a curiosity. "Four hundred people today, peaking 6–7pm" is a staffing decision.
Use case 03
Restricted zone intrusion
A maintenance vehicle enters a pedestrian-only walkway. The panel turns red the moment it crosses in.
Nobody labelled that truck. You define the zone; Arjuna works out the rest. The same rule covers a fire lane, a loading bay after hours, or a substation perimeter.
Use case 04
Loitering & dwell time
A dwell timer runs on every tracked person; the alert fires past the threshold.
Duration is the signal a human monitoring forty screens cannot track. The threshold is yours to set six seconds in a doorway, two minutes in an ATM lobby.
Use case 05
Crowd density
Dense, overlapping crowds in a public concourse.
The hard case: people behind people. Density and flow, measured continuously.
Use case 06
Retail footfall
Entrance and escalator counting.
Conversion needs a denominator. Arjuna gives you the number of people who walked in, not just the number who bought.
Use case 07
Vehicle detection & classification
Vehicles detected and split by class car, truck, bus.
Not a single blob count. The difference between "traffic was busy" and "commercial vehicle share rose 12% this quarter."
Use case 08
Parking occupancy
Aerial view of a dense lot, vehicles counted and classified.
Occupancy without a single ground sensor. Retrofit to a camera you already own instead of trenching the car park to install loops.
Use case 09
Low-light performance
Night street scene, vehicles and signals detected.
Low light is where cheap analytics quietly stop working. Here's ours, unedited, so you can judge it yourself.
Safety compliance
The differentiator
Detection tells you what is in frame. Compliance tells you whether your people are working the way your safety policy requires. Arjuna uses pose estimation the position of shoulders, wrists and hips to judge behaviour, not just presence.
Compliance 01
Handrail compliance
Wrist keypoints are measured against the handrail, calibrated once per camera. Every person on the stairs is scored: holding, or not holding.
A safety policy that nobody measures is a policy nobody follows. Arjuna gives you the compliance rate for a staircase, per shift, as a number.
Compliance 02
Lab coat compliance & doorway entry
Garment check on the torso region, plus entry detection at a defined doorway.
Who entered the controlled area, when, and were they gowned correctly. Logged automatically, every time.
Deployment & privacy
Runs on your infrastructure
Footage never leaves your premises. On-premise by default.
Existing cameras
Standard RTSP. No hardware replacement.
Redaction built in
Faces and plates blurred at the point of processing.
DPDP-aligned
On-premise processing and redaction as defaults, not add-ons.
What we tell you honestly
A page that claims everything works perfectly reads as marketing. Naming limits reads as engineering.
Camera placement decides what is possible.
Some checks need an angle you may not have yet. Eye protection, for example, cannot be verified from a camera mounted down a corridor the face is too small, and ordinary spectacles are indistinguishable from safety goggles at that distance. We will tell you where a camera needs to move rather than ship a check that quietly passes an unprotected worker.
Fixed mounts matter.
Zone and handrail rules are calibrated to a camera's view. A camera that shifts needs recalibration. Fixed CCTV is stable by design; this is one of the few places where existing infrastructure beats a phone.
Custom checks need custom training.
Generic detection people, vehicles, zones works out of the box. Site-specific classes such as a particular uniform, badge or piece of equipment need a short training cycle on labelled examples from your site. Typically about a week per class.
We start with a site assessment.
Before quoting, we look at your actual camera angles and tell you which use cases your current estate supports today, and which need a camera moved.
See Arjuna on your own cameras.
Send us a short clip from one camera. We'll run it and return an annotated result, so you can judge the system on your footage rather than ours.