Smart Retail: Edge AI People Counting 
Real-Time ROI Analytics | High-Performance Low-Power Vision AI
 
Transform retail foot traffic into actionable data. Our demo features ROI-based counting and dwell-time analytics running entirely at the edge on the Grinn GenioBoard - zero-latency insights with 100% on-device privacy.

The Efficiency-First Edge AI Stack

CrossPlay Integration: A versatile streaming and interaction stack that brings desktop-class connectivity to embedded HMIs. It enables seamless WebRTC remote interaction and real-time Vision AI overlays within a unified Qt-based architecture.

Grinn GenioBoard (Genio 700): Built on an advanced 6nm process, this octa-core SoC delivers "desktop-class" performance with outstanding thermal efficiency. Optimized for 24/7 operation, the board features a typical power consumption of just ~6W, making it ideal for fanless, energy-efficient retail deployments. While our 42" demo runs in Full-HD, the board’s HDMI and DisplayPort outputs natively support up to 4K60.

DeepX M1 NPU Acceleration: By offloading AI workloads to the M.2 PCIe accelerator, we achieve elite-level throughput (25 TOPS) at a mere 3W-5W. This allows the CPU to stay cool and responsive, ensuring a lag-free Full-HD touch interface even during peak traffic.

Ready for EU CRA Compliance: Built with security in mind, featuring the Infineon OPTIGA™ Trust M module. This provides hardware-based key storage and cryptographic integrity, simplifying the path to meeting Cyber Resilience Act (CRA) requirements for IoT devices.
 

Core Demo Features

ROI-Based Counting: Define specific "Power Zones" to track engagement with premium store displays.

Dwell-time Analysis: Quantify how long customers interact with products to optimize floor layouts.

Bi-Directional Flow: Accurate entry/exit tracking to calculate real-time conversion rates.

Privacy by Design: No video data leaves the device. Local processing ensures 100% data sovereignty.

Modern Interactive UI: Experience a lag-free, high-refresh dashboard with real-time telemetry overlays. Effortlessly manage multitasking and touch interactions without impacting AI detection performance.

• Type: AI Accelerator​

• Form Factor: M.2 M Key (22 x 80 mm)​

• Interface: PCle Gen.3 x4​

• Memory: 4GB LPDDR5, QSPI 1Gbit NAND Flash​

• Host HW: x86, ARM Based Architecture