AI-Powered Real-Time Quality Control
summary
An AI vision pipeline deployed in Docker on edge devices and Azure integrates with PLCs and IoT hubs, using Kean/Cognex cameras, background‑removal plus defect‑classification models (ResNet family) and OOD detection to run inference every ~0.5s. It cuts scrap by 90% within 50 days, reaches ~100% precision and recall after iterative retraining, and scales toward 40 lines with a projected 1.2 million euro annual savings. Production issues addressed: lighting, polarization, data throughput, firewall/OT network integration.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:30Computer vision in manufacturing
- 6:44Problem statement and manual QC
- 10:59AI-powered quality control
- 13:11System architecture and training
- 23:41Defect detection and performance
- 26:31Business impact and savings
- 27:54Future scaling and improvements
- 29:39Q&A
related talks
10mRealizing the Power of AI at the Edge
17mAI in Quality Control – Real Use Cases and Best Practices
27mComputer vision for automated quality inspection
06mAI in production: The droids you’re looking for
59mAI in the Battle Against Fakes
43mIntersection of Awesome: Building AI enabled IoT solutions powered by the cloud
01h 02mAI on the Edge
45mFrom Data to AI on The Edge