Intelligence
at the edge.
Crossware integrates Edge AI into production embedded products — from driver monitoring and sign recognition in automotive, to people counting in industrial environments. No cloud dependency. Real hardware. Shipped software.
AI inference that runs where your product runs.
Edge AI in embedded products is not a research problem — it's an integration problem. Crossware handles model selection, quantisation, hardware-accelerated deployment and HMI integration on the same platforms your product ships on.
All implementations shown on this page run on actual hardware at production frame rates. No simulations, no desktop proxies.
Drowsiness and fatigue detection in real time.
Crossware's DMS implementation tracks eye aspect ratio (EAR), mouth aspect ratio (MAR), head pose, gaze direction and blink frequency to detect drowsiness and fatigue. Pose estimation adds posture monitoring and hands-on-wheel detection. All inference runs on-device at production frame rates — no cloud, no latency overhead.
From monitoring to active alerting.
The system outputs structured alert states — drowsy, distracted, eyes-closed, head-down — that integrate directly with the vehicle HMI and safety systems. Crossware engineers the full pipeline: from camera input and model inference to alert classification logic and Qt-based HMI integration, all running on the same embedded SoC.
Traffic sign detection and classification.
Crossware implements real-time traffic sign detection and classification on embedded automotive platforms. The vision pipeline handles sign localisation, classification and confidence scoring at production frame rates, with results surfaced directly to the HMI. Speed limit signs, stop signs and regulatory signage are classified and passed to the cluster or safety system.
Occupancy and movement tracking on-device.
Crossware has deployed people counting and zone occupancy monitoring in industrial facilities — using overhead camera feeds and on-device inference to detect, track and count people across defined zones without any cloud connectivity. Output integrates with building management systems, access control and safety monitoring applications.
Dedicated edge-AI accelerator purpose-built for embedded inference. Crossware integrates DeepX as a co-processor alongside application SoCs for high-throughput vision workloads.
Integrated NPU with eIQ ML software stack. Crossware targets the i.MX8M Plus for DMS and industrial AI on the same SoC as the product HMI — no separate AI board required.
Next-generation platform with enhanced NPU performance for automotive-grade ADAS and DMS workloads. Crossware supports full BSP, Yocto and eIQ integration for i.MX95 targets.
Where Crossware has shipped Edge AI.
Driver Monitoring Systems
Production DMS integration covering the full pipeline from camera to HMI alert — deployed on i.MX8M Plus and i.MX95.
- Drowsiness and fatigue detection
- Eye closure (PERCLOS) and blink monitoring
- Head pose and gaze direction
- Yawn detection
- Posture and hands-on-wheel monitoring
- Traffic sign recognition
People Counting & Safety
On-premise occupancy monitoring without cloud dependency — suitable for restricted access areas, cleanrooms and safety-critical zones.
- Zone-based people counting
- Entry and exit tracking
- Occupancy alerts for safety systems
- Integration with building management systems
- Privacy-compliant on-device processing
Bring AI inference to your embedded product.
Share your target hardware, camera setup and detection requirements. Crossware will propose the right model architecture, NPU integration path and HMI implementation for your product.