CrowdSense-AI

Real-Time Crowd Counting & Analytics

Entry & Exit Detection with Edge-Accelerated AI Inference

24/7 Support

Request a Demo → ← All Products

Key Capabilities:

Head Counting:

Line-crossing counter with entry/exit detection using single counting line
Cross-product direction analysis automatically determines in or out
Anti-jitter logic with configurable cooldown frames, minimum displacement and deadzone
Per-event snapshot capture — automatic cropped snapshot of each detected person
Deduplication by distance and frame window prevents double-counting
Configurable confirm frames for entry and exit events separately

Edge AI Inference:

Hardware-accelerated inference on edge AI hardware at 12+ frames per second
Dedicated head detection model optimised for accuracy in crowded scenes
Detects heads reliably under partial occlusion, varied lighting and different viewing angles
Head-specific model — purpose-built for bus passenger counting scenarios
Optimised for edge AI hardware acceleration
95%+ detection accuracy in real-world conditions

Multi-Head Tracking:

Advanced tracking algorithm maintains unique identity per detected head across frames
Handles occlusions, overlapping paths and re-identification seamlessly
Smooth trajectory prediction across frames
Configurable tracking thresholds for matching and lost track buffer
Centroid-based maximum distance for accurate person-to-track assignment
Minimum track frames setting to filter out noise detections

Live Dashboard & Analytics:

Real-time video feed with head bounding boxes, person IDs and counting line overlay
Live counters for In, Out and Total updated instantly via WebSocket
Interactive charts — in vs out counts, hourly activity distribution
Recent event panel with snapshot previews and direction indicators
FPS, uptime and system status indicators always visible
Accessible from any device on the network via web browser

Reports & Documentation:

PDF reports with summary statistics, event breakdowns and formatted tables
Word (.docx) reports for editable deliverables and team annotations
CSV export for further analysis in Excel, Power BI or custom tools
Custom date range selection for any reporting period
Built-in user manual with screenshots, exportable as PDF or Word
On-demand report generation with one-click download

Video Recording & Playback:

Record camera feed to disk continuously with configurable settings
Browse and playback recordings organised by day and hour
Delete individual recordings for storage management
Video Center page with date-based navigation
RTSP and file source support for IP cameras and testing
Continuous recording without impacting detection performance

How It Works:

Capture — Video frames captured from RTSP camera or video file at configured frame rate.
Detect — Hardware-accelerated AI model detects all heads within the Region of Interest.
Track — Tracker assigns and maintains a unique ID for each detected head across consecutive frames.
Count — When a tracked person crosses the counting line, direction analysis determines in or out with anti-jitter protections.
Record & Report — Events stored in database, snapshots saved, video recorded. Generate reports in PDF or Word format.

Feature Details:

Detection & Tracking:

Head detection model optimised for edge AI hardware acceleration
Multi-object tracker with smooth trajectory prediction
Configurable detection confidence threshold and filtering parameters
Region of Interest — drag-to-draw ROI rectangle to limit detection area
Minimum detection area filtering to ignore distant/small objects
Frame skip support for performance vs accuracy trade-off

Platform & Infrastructure:

Hardware: compact, fanless edge AI device
OS: Linux-based embedded operating system
Database: secure relational database for events and configuration
Streaming: live video and event streaming over standard web protocols
Frontend: responsive web dashboard with interactive analytics charts

Counting Engine:

Single counting line with cross-product direction analysis
Configurable deadzone, cooldown and deduplication parameters
Multi-frame confirmation for reliable entry and exit detection
Direction swap support for flexible camera installation
Minimum confirm distance and track frame thresholds for noise filtering
Optional image preprocessing for challenging lighting conditions

Licensing & Security:

Cryptographically signed license files bound to device hardware ID
Unique device fingerprint prevents unauthorised license copying
Per-feature gating — individual modules licensed independently
License status with colour-coded expiry warnings
Role-based access — Admin (full control) and User (view + reports only)
Session-based authentication with comprehensive REST API

Configuration & Customization:

Video Sources — Add, edit, and switch between multiple RTSP cameras or video files.
Detection Tuning — Adjust confidence threshold and frame skip to balance accuracy vs performance.
Region of Interest — Drag-to-draw ROI rectangle to limit detection to a specific area.
Counting Line — Position the counting line by dragging on live preview. Direction of crossing determines in or out.
Overlay Controls — Toggle overlay items: in/out counts, FPS, bounding boxes, person IDs, counting line, ROI rectangle.
Event Snapshots — Automatically capture and store cropped snapshot images of each detected person.

Deployment & Licensing:

Installation & Setup:

Automated setup installs all dependencies and configures the system
Provisions the runtime environment automatically
Installs as a managed service for auto-start on boot
Pre-configured edge AI runtime environment ready out of the box
Supports RTSP IP cameras and video files for testing
Full setup from unboxing to first detection in under an hour

Web Interface (16 Pages):

Dashboard with live video feed, counters, FPS and event timeline
Camera configuration with ROI drawing directly on live canvas preview
Configuration for detection thresholds, counting line position and overlay toggles
Session management for bus/route tracking
Employee management, accounts and role-based access control
Reports, recordings, logs, network config, system info, license and user manual

What You Need / What We Provide:

Hardware: compact edge AI device
Camera: RTSP-compatible IP camera or video file for testing
Network: local Ethernet or Wi-Fi connection
Browser: any modern web browser (Chrome, Firefox, Edge, Safari)
We provide: pre-configured edge device, pre-installed CrowdSense-AI software, installation assistance, ROI setup, operator training and software updates during license period

API & Integration:

Full REST API covering authentication, head counting, video sources, reports, recordings, charts and system management
WebSocket for real-time event notifications and live video frames
Comprehensive API endpoints for full programmatic control
Secure session authentication with encrypted credentials
MJPEG video feed endpoint for embedding in external monitoring tools
Health endpoints for external system polling and integration

CrowdSense-AI

Key Capabilities:

Head Counting:

Edge AI Inference:

Multi-Head Tracking:

Live Dashboard & Analytics:

Reports & Documentation:

Video Recording & Playback:

How It Works:

Feature Details:

Detection & Tracking:

Platform & Infrastructure:

Counting Engine:

Licensing & Security:

Configuration & Customization:

Deployment & Licensing:

Installation & Setup:

Web Interface (16 Pages):

What You Need / What We Provide:

API & Integration:

Product Walkthrough: