Deep-learning in industry: design of a Convolutional Neural Network for quality inspection in mass production.

The core of this thesis is the use of Artificial Intelligence for quality inspection purposes. The quality control system devised based on deep-learning deploys in particular the convolutional neural network structure. The challenging idea was to design a CNN from scratch able to recognize images of defective caps that could replace the already implemented Image processing system. Chapter 2 contains a comparison between those two different technologies, focusing in particular on the strengths that lead to opting for the Convolutional Neural Network structure to replace the inspection system currently run by the automatic machines. Chapter 3 contains all the theory necessary to understand the Convolutional Neural Network working principle, starting from the basic concept behind neural networks to the detailed explanation of all the layers constituting the used network. Finally, chapters 4 and 5 explain all the code written to develop the project. In particular, chapter 4 explains all the pre-processing steps necessary to enhance the learning step whereas chapter 5 reports all the results obtained through the simulations.