Using deep learning techniques for automated inspections
Products manufactured by Prodrive Technologies range from smart thermostats and automotive wallboxes to large compute cabinets and gradient amplifiers. To ensure high-quality production, all these products have a final, visual inspection step where each product is inspected before it is packed and shipped to customers. Inspections including presence and anomaly detection are automated using the Keras and Tensorflow deep learning libraries where the classification and autoencoders models are developed and trained in-house. To improve the factory efficiency even more, these techniques will also be applied to factory data such as cycle times and process parameters for the purpose of diagnostics and predictive analysis.