Georgios Koutroulis, researcher from Area 3 – Cognitive Decision Systems, presented his work at the 12th International Conference on Knowledge Discovery and Information Retrieval (KDIR 2020). The KDIR 2020 conference focused on methodologies for identifying valid, novel, potentially useful and meaningful patterns from data, often based on underlying large data sets.
The presented paper, titled “Enhanced Active Learning of Convolutional Neural Networks: A Case Study for Defect Classification in the Semiconductor Industry”, depicted the development and the evaluation of a novel algorithmic framework that classifies images of defected dies from silicon wafers. Our active learning approach alleviate the burden from the process engineer to label thousands of defect images, as it selectively picks the batch of images that will help the deep learning model to learn faster and better.
This paper was co-authored with Tiago Santos from TU Graz and supervised by Roman Kern and Stefan Thalmann along with support from TDK Electronics. The conference proceedings are yet to be published. They will soon be published at the SciTepress Digital Library and Springer.
With this publication, the project DEFCLAS (Automatic Defect Classification in Silicon Wafers) focusing on high-performance defect classification systems on silicon wafers marks the end of the project.
#deeplearning #defectclassification #semiconductorwafer #activelearning #pro2future #cognitive
