People Analytics

Smart Tracking System

AI-powered smart video tracking system for multi-camera person search using 50+ appearance attributes. Fast, privacy-aware - ideal for Safe City, retail, and transport.

Real-time inference
On-premise or cloud
GDPR compliant
REST API & SDK
CAM-01 Entrance
REC
CAM-04 Corridor B+3 min
CAM-07 Parking+8 min
MOVEMENT PATH
Re-ID Attributes
  • Hat: Red cap 98%
  • Top: Red jacket 96%
  • Bag: Brown sling 91%
  • Pants: Dark 89%
  • Gender: Male 94%
14:02CAM-01
14:05CAM-04
14:10CAM-07
50+
Attributes to search by a person's appearance
Process

How It Works

01

Ingest & Detect

Cameras stream video into the system. An AI model detects humans, extracts appearance embeddings (a compact numerical signature of clothing, bag, posture) and timestamps each detection.

02

Attribute Extraction

Each detection is labeled with attributes: gender, age group, clothing color, bag presence, hat, carrying objects, gait clues, and more. These attributes create the searchable profile.

03

Cross-Camera Association

sing a combination of appearance embeddings, trajectory modeling, and timestamps, the system links detections across cameras to build a continuous route - even when the individual leaves one camera’s view and reappears on another.

04

Search & Playback

Operators type or select attributes and get ranked results - short video snippets showing candidates across cameras, plus a reconstructed timeline and route map.

Integrations

Works With Your Existing Stack

Deploy alongside your current cameras, hardware, and VMS - no forklift upgrade required.

IP Cameras

Any ONVIF / RTSP stream

Axis CommunicationsBosch Security SystemsDahua TechnologyHanwha VisionHikvisionVivotekUniview (UNV)MilesightRTSPONVIF

VMS Platforms

Native plugin & SDK support

VMS VezhaMilestoneGenetecNx WitnessAvigilon

Edge Hardware

GPU-accelerated inference

NVIDIA JetsonIntel NUCGPU ServerCloud
FAQ

Frequently Asked Questions

Contact Us

Ready to Get Started?

Fill in the form and our team will get back to you shortly.