We demonstrate, test, and benchmark an automated robot assisted machine vision cell using unsupervised AI anomaly detection for zero defect manufacturing quality control. A robot arm manipulates complex parts across multiple camera angles and dynamic lighting setups, enabling automated OK or NOT OK defect classification without requiring large defect image training sets. The purpose is to help discrete manufacturing companies inspect complex 3D geometries, reducing false rejection rates by up to 30 percent and eliminating manual inspection bottlenecks.
A documented demonstration of robot-assisted visual quality inspection using AI anomaly detection, including test observations, classification performance, and recommendations for further development toward production-ready inspection applications. This setup reduces quality control deployment time by up to 50 percent since models are trained solely on good parts, achieving over 98 percent defect detection accuracy across proven trial runs in precision metal casting and plastic injection molding.
The service is delivered as a test and demonstration process including setup of a robot-assisted vision cell, controlled presentation of parts for image capture, execution of AI-based anomaly detection, assessment of OK/NOT OK classification performance, and delivery of documented findings and recommendations for future production-ready inspection solutions.
Manufacturing, Quality control and inspection, Robotics and automation, Machine vision, Industrial product and process validation