The situation
The electronics factory relied on manual visual inspection, which was slow, inconsistent, and let defects through to later stages — driving up scrap and rework costs.
What we did
We built a computer-vision system using PyTorch and YOLOv8 with OpenCV, deployed via ONNX Runtime and served through FastAPI for real-time inspection on the line. Grafana dashboards track detection rates and defect trends.
The outcome
The system detects 99.7% of defects, scrap fell by 60%, and the investment paid for itself within 4 months.