Watchman project, Lombardy

WATCHMAN (Workload-reduction machine vision-based technology Hub for manufacturing) project aims to develop a hub of expertise and experimentation on the application of Artificial Intelligence techniques in Machine Vision in manufacturing. In particular, two use-cases has been developed, the first one concerning pharma industry and the second related to automotive sector:

• SALF demonstrator: creation of a flexible and easily configurable system capable of detecting print defects and specific parts in bags of saline. The challenges are related to different shapes of the saline bags, the hard-to-identify printed text, the reflection of the liquid inside the bag and other issues related to the presence of bubbles and other artifacts inside the bag. The developed AI solution is related to machine vision and image analysis, in order to analyse the shape and presence of all the parts that make up the bag and to verify the correctness and legibility of the printed information.
• BREMBO demonstrator: detection of paint or mechanical defects at the end of the painting or oxidation process of the brake callipers to ensure the quality of the products. The goal is to create a flexible and easily configurable computer vision system capable of detecting surface defects, or painting brake callipers. The developed solution should face several challenges, such as the high geometric, dimensional and colour variability, the different types of defects and the emergence of possible problems related to the development of new parts.

The solution is based on two approaches of deep learning techniques: a semi-supervised machine learning detection algorithm and the Few Shot Learning model that allows to classify different types of anomalies using a small amount of images in the initial training phase.