AI Video Analytics Platform
Guard EkMitra
Turn passive video into searchable, actionable and quantifiable intelligence. Guard EkMitra extracts metadata from any camera feed, so every person, vehicle, face and object becomes a record you can search, and every rule becomes a real-time alert.
- Licence plate recognition (ANPR) for Indian plates
- Person, vehicle, face and object attributes
- Line crossing, loitering, crowd and zone rules
- Watchlist match to alert in under 5 seconds
- Works with existing cameras, VMS and NVR
- Edge AI: only metadata leaves the site, never video
Objects
4 tracked
Inference
30 FPS · Edge
Alerts
0 open
Edge AI on the camera network. Lightweight metadata to the centre.
Existing cameras and video systems stay as they are. AI runs on edge nodes next to the cameras, only text-sized rows travel to the centre, and alerts reach the control room in under five seconds.
Searchable
Every object, attribute and event becomes a row you can filter and find in seconds.
Actionable
Rules turn detections into real-time alerts worth a human response.
Quantifiable
Counts, dwell, occupancy and trends become dashboards and reports.
- Government camera grids (RTSP / WHEP / HLS)
- Department VMS / NVR
- Uploaded footage (MP4 / MOV)
Connects to camera grids, VMS and NVR feeds. Samples frames behind a motion gate, then detects, tracks, reads plates and captures attributes.
One structured record per object with a best-shot crop. Only text-sized rows travel to the centre, never video.
Zones, lines, loitering, dwell, watchlist hits and camera offline. Sighting to alert in under 5 seconds.
Live wall, camera health, GIS map with route trace, plates and watchlist, events, alerts and insights KPIs.
Three capability layers · Person · Vehicle · Face · Object
LAYER 1
Object & vehicle recognition
- Licence plate recognition (LPR / ANPR)
- Vehicle classification
- Bag or weapon detection
- Package tampering
- Object left or removed
- Facial recognition
LAYER 2
Behavioural analytics
- Line crossing & perimeter breach
- Loitering & restricted zones
- People counting & occupancy
- Crowd density & queues
- Wrong-way, speed & direction
- Slip & fall, anomaly detection
LAYER 3
Advanced video analytics
- Multi-camera object tracking
- Time compression for fast review
- Weapon detection
- Emotion recognition
- Anomaly detection
- Camera vision
One console for every camera, event and alert.
Every tile shows the value, its change against the last period and a sparkline, and drills down to the evidence. Dark by default for the control room, with a light theme one click away.
- Insights
- Live wall
- Cameras & health
- Objects
- Events
- Faces
- Recordings
- Rules
- Alerts
- GIS map & route trace
Vehicles read
ANPR reads, unique plates and read rate
Hotlist hits
Vehicle and face watchlist matches, confirmed vs pending review
Alerts
Raised, acknowledged, resolved and median time to acknowledge
False alarm rate
Share of alerts operators mark false, per rule
Cameras online
Online of total, with cameras offline for over an hour
Persons detected
Counts with the night-time share for after-hours safety
Operator workflow
STEP 1
Onboard
Import the camera catalogue into the registry and map
STEP 2
Connect
Pull streams over the grid or VMS, reconnect automatically
STEP 3
Sample
Frames behind a motion gate, so idle cameras cost nothing
STEP 4
Analyse
Detect, track and read plates and attributes
STEP 5
Match & alert
Watchlists and rules, alert in under 5 seconds
STEP 6
Investigate
Triage alerts, then trace a plate's route on the map
Industries & use cases
Corporate offices
Work-life balanceSee how time splits between work areas, cafeteria and recreation zones, and the gym. Spot teams that are overworked, plan breaks and wellness programmes, and track meeting-room and desk occupancy.
- Work, fun and fitness time by floor or team
- Late-hour presence for employee safety
- Meeting-room and desk utilisation
Workplace insights are shown as zone and team-level totals, with employee notice and consent, in line with India's DPDP Act. Individual tracking is switched off unless your policy requires it.
Casinos
Gaming floorWatch the floor, the cash cage and the entrances from one console.
- Watchlist match for banned or self-excluded guests
- Table and floor occupancy
- Cash-cage perimeter and loitering alerts
Government & public safety
Control roomsDistrict edge nodes on existing camera grids, with plate watchlists and route tracing.
- ANPR and hotlist alerts
- Route trace on the GIS map
- Crowd and perimeter rules
Smart cities & traffic
RoadsVehicle counting, classification and wrong-way detection at junctions.
- Traffic counts and speed
- Wrong-way and violation alerts
- Rider-without-helmet rules
Manufacturing & warehouses
PlantsKeep people safe and assets where they belong.
- PPE and restricted-zone alerts
- Object left or removed
- Slip & fall detection
Retail & malls
StoresUnderstand footfall and keep queues short.
- Footfall and dwell time
- Queue length alerts
- Theft and loitering alerts
Hospitals
HealthcareProtect patients and staff across wards and entrances.
- Restricted-area access
- Slip & fall alerts
- Visitor and crowd monitoring
Schools & campuses
EducationSafer gates, hostels and grounds.
- Gate entry by plate or face
- After-hours perimeter alerts
- Crowd density at events
Hotels
HospitalityDiscreet security that does not disturb guests.
- Parking ANPR and VIP arrival
- Unattended baggage in lobbies
- Back-of-house access alerts
Public transport
StationsStations, depots and platforms under one view.
- Platform crowd density
- Unattended baggage
- Track and perimeter intrusion
Three imperfect reads, one vehicle.
one key · one sighting history · one route
Misreads fold into one key, so a watchlisted vehicle is still caught when the camera misreads a character, and every sighting joins one route on the map.
Why a platform, not a point product
On-camera analytics
- Limited processing power
- More false alarms
VMS-embedded analytics
- Limited capabilities and scale
- Tied to one vendor
Point products
- One use case each: faces, plates or counting
Guard EkMitra platform
- Every use case on one console
- Scales from one site to a whole state
- Higher accuracy, faster review, business intelligence
Automated review
Time compression turns hours of footage into seconds of search.
Fewer false alarms
Staff attention goes to real events, not noise.
Proactive alerting
Prevention instead of after-the-fact review.
Data-driven operations
Business intelligence from cameras you already own.
- Human oversight of every alert
- Transparent about what is detected and why
- Work to remove bias
- Privacy by design: metadata, not video
- Data integrity and access control
- Every alert and verdict audited
Metadata only
Rows, never video, cross the network.
Scale out per district
Identical edge nodes, one centre.
Auditable
Every alert verdict is logged.
Open
REST + OpenAPI, WebSocket alert stream, metrics.
Integration-ready via REST (roadmap): VAHAN owner by plate, SARTHI licence, eGujCop FIR / hotlist to watchlist, AFIS / NAFIS identity.
See everything, understand anything, act on what matters.
See Guard EkMitra on your own cameras.
Tell us how many cameras you have and what you want to watch for. We'll plan a short pilot.