AI Inspection
ONNX-based inference on the live feed
Run object detection, classification, and anomaly detection models directly on the camera stream. GPU acceleration, custom model management, and real-time overlays bring AI-powered quality control to your inspection workflow.
Capabilities
AI-powered defect detection
Optysys integrates ONNX models for real-time inference on the live video feed. Upload custom models, configure labels, and visualize detections as overlays — all without leaving the inspection interface.
- GPU AccelerationLeverage CUDA or DirectML for fast inference. ONNX Runtime automatically selects the best available execution provider.
- Model ManagementRegister ONNX models in the web portal. Configure input dimensions, labels, and model type per project.
- Real-Time OverlayDetection boxes and labels rendered directly on the live feed. Toggle overlays on or off during inspection.
- Custom ModelsUpload your own ONNX models. Support for object detection, classification, segmentation, and anomaly detection.
- Component DetectionDetect and classify PCB components, solder joints, and assembly defects with trained object detection models.
- Anomaly DetectionIdentify unusual patterns and defects using anomaly detection models trained on normal samples.
From model to inspection
Create or upload ONNX models in the web portal, assign them to projects, and enable inference in the desktop app. The live feed runs through the model pipeline in the background — detections appear as overlays while you inspect.