Artificial Intelligence, particularly Deep Learning models, has shown remarkable potential in revolutionizing industrial automation. This thesis focuses on leveraging such models, specifically Convolutional Autoencoders, for defect detection in industrial components. The research addresses the automation of quality control in brake calipers within a typical automotive production line setting. The primary objective is to develop an automatic defect detection system capable of identifying various imperfections, such as deformities, bubbles and scratches, on brake calipers. A Convolutional Autoencoder is trained to decode input images of brake calipers, reconstructing them to remove potential defects. A difference function, accounting for contrast, brightness and structure, is then applied between the original and reconstructed images to identify candidate defective areas.