Defect Classification in Semiconductor Manufacturing Images: A Novel Weakly Supervised Approach

In recent years, automatic anomaly detection techniques have been widely used in industrial manufacturing. In this thesis, the aim was to classify a specific type of defect (corrosion) in a dataset related to semiconductor manufacturing. Due to the limited amount of images available, I approached the problem in a semi-supervised environment using common anomaly detection methods. In the end, I came up with a completely new method for classifying defects in images.