Dipartimento di Ingegneria informatica, automatica e gestionale

This paper reports on the design and development of a customized Automated Optical Inspection (AOI) solution aimed at detecting defects in a production line related to the correct mounting of integrated circuits. Contrary to most solutions avail- able on the market, the developed system relies on deep learning to be able to perform detailed real-time visual inspections of components without the need to compare the captured photos with any reference images/golden sample. The proposed solution was designed to also provide good generalization capabilities, accommodating visual changes in the environment and in the structure of the component being produced. A custom testing machine was built in order to perform real-time inferences and validate the simulation results in a real-world setting.