PRODUCTS / AI BOX SERIES / SMART RETAIL COUNTER
Smart Retail Counter – People Counting and Queue Analytics Camera System
A compact unit in the back office that turns existing store cameras into footfall, queue, and dwell measurement. One unit per store, no server, no IT visit.
Counts people. Does not identify them. No face templates are created by this module.
What It Measures
Entry and exit counts
Demographics: Gender and Age Estimation
Queue length and wait time
Dwell time by zone
Heatmaps of floor movement
Real-Time In-Store Analytics
People Counting
- Accurate entry and exit counting
- Live visitor statistics
- Daily and hourly summaries
Queue & Congestion Monitoring
- Queue length detection
- Wait-time alerts
- Congestion hotspots
Heatmap Visualization
- Customer movement heatmaps
- High-traffic zone identification
Why Choose the AI Box Edition
Best suited for retailers who want immediate insights without IT complexity.
No existing servers required
Works directly with existing IP CCTV cameras
Edge processing with no cloud dependency
HOW IT WORKS
Deployment Steps
One Unit per Store
Sits in the back office. Connects to the cameras already covering the entrance and the floor.
Draw Lines and Zones
Counting lines at entrances, zones on the floor, queue areas at the till.
Read The Numbers
Local dashboard per store. Export or push upstream for chain-level reporting.
Flexible Hardware for Any Scale
The ARSA AI Box is available in two distinct hardware configurations, allowing you to choose the architecture that best fits your facility’s infrastructure, whether distributed at the edge or centralized in the server room.
The Edge Model (Mini PC)
Process up to 3 Camera
Ideal for distributed deployments and space-constrained environments.
- Core Architecture: Built on the Intel® NUC platform.
- Form Factor: Compact, ultra-small footprint designed for easy mounting behind displays or in tight spaces.
- Performance: Optimized for low-power, continuously running 1-3 streams or low-density inference tasks.
- Deployment Strategy: Place one unit per location to keep processing localized and minimize bandwidth usage.
The Server Model (High-Density)
Process up to 25 Camera
Ideal for centralized monitoring and heavy workload consolidation.
- Core Architecture: Powered by Intel® Processors paired with dedicated NVIDIA® GPUs.
- Form Factor: Industrial rack-mount chassis for seamless integration into standard server cabinets.
- Performance: High-throughput processing capable of analyzing up to 25 camera streams simultaneously in real-time.
- Deployment Strategy: Route all camera feeds to a single, powerful central unit to simplify hardware management and maintenance.
TRANSPARENT
Pricing
The Appliance Hardware
| System | Cameras | Price |
|---|---|---|
| AI Box Mini | up to 3 | $1,890 |
| AI Box Server | up to 25 | $10,900 |
| AI Box Server Pro | 25+ | Request a quote |
Retail Counter Licence, per camera
| Module | Perpetual | Maintenance / year | Subscription |
|---|---|---|---|
| Retail Counter | $870 | $157 | $29 / month |
The appliance is purchased outright and includes the configured unit, local dashboard, remote setup, and a 12-month hardware warranty. Licences are bought per camera, outright or by subscription. Camera hardware is not included.
MINIMUM
Site Requirements
| Overhead versus angled mounting | Line counting is optimal at 60° to 90° (near-nadir). Queue and dwell require 30° to 50° oblique mounting for occlusion control. |
| Minimum resolution at measurement distance | 113 PPM for entry/exit line counting. 99 PPM for queue length and dwell measurement. |
| Accuracy in dense crowds | High at near-nadir angles for entry counting. Oblique angles below 30° result in under-counting due to occlusion. |
| Lighting tolerance | Minimum 50 lux at target. Consistent indoor lighting without harsh dynamic shadows recommended. |
FAQ
Common Questions
Should cameras be overhead or angled?
Overhead (60 to 90 degrees) is optimal for entry and exit line counting to prevent occlusion. Angled or oblique (30 to 50 degrees) is required for queue and dwell measurement so the camera can see individual heads separated in the queue.
How accurate is counting in dense crowds?
Accuracy is maintained if the camera is mounted steeply overhead (near-nadir) for counting. At shallower angles (below 30 degrees), the queue collapses into a single blob in the image, causing under-counting.
Can one unit cover multiple entrances?
The Mini covers up to three camera streams, enough for most single-store deployments with a main and a service entrance. Larger stores or supermarkets use the Server.
Does it identify individual shoppers?
No. This module counts and tracks anonymous shapes. It does not create face templates and cannot recognise a returning individual. Aggregate data is entirely anonymous.
How do I roll this out across 40 stores?
One Mini per store on subscription, shipped configured. Rollout is a courier delivery and a 15-minute connection per site, not an engineering visit. The units will immediately begin reporting into a single chain-level dashboard.
Turn Store Footage into Retail Intelligence
Understand customer behavior and optimize store performance in real time.