Computer Vision Industrial Automation

Edge AI pipeline using multispectral cameras to detect defective parts on a manufacturing line and automate sorting in real time — no network dependency, sub-millisecond latency.

This industrial AI project deploys a computer vision pipeline directly on a manufacturing production line. Multispectral cameras (visible, infrared and UV) capture high-resolution images of components as they move along the conveyor; a custom-trained machine learning model analyses each frame in real time to detect defects, surface anomalies and non-conforming geometry.

The system runs entirely on edge hardware — no cloud dependency — ensuring sub-millisecond reaction times even in environments with no network connectivity. When a defect is detected, the pipeline instantly triggers pneumatic deflectors or robotic arms to divert the faulty part before it advances to the next production stage.

A real-time monitoring dashboard gives operators full visibility into line throughput, defect rates by category, and a complete traceability log of every quality event — enabling continuous process improvement and audit-ready reporting.