Quantified Automated Optical Inspection

Quantified Automated Optical Inspection: Fully enabling machine learning for enhanced hardware assurance outlines various inspection modalities, the growing popularity of machine learning (ML) techniques for enhanced AOI, critical bottlenecks in these approaches, and solutions that increase scalability, reliability, and accuracy of ML-based AOI techniques. As a result, readers will clearly see which trends must be updated to properly incorporate deep learning strategies given the constraints present in PCB analysis. These constraints include limited publicly available data, lack of rigor/consistency in ground truth collection, lack of result quantifiability, and limited diversity in fundamental ML approaches. Solution spaces will be explored such as enhanced generative models to quantify PCB component properties, verification against component datasheets, and codified recommendations for rigorous data aggregation and dissemination.