Abstract
Wafer is a thin slice of semiconductor substance used for fabricating integrated circuits in semiconductor manufacturing. Wafer Bin Maps (WBM) are the results of Circuit Probe inspection of the dices on the wafer, which provide crucial information to identify the root cause of the problems in semiconductor manufacturing. Automatic identification of defect patterns in WBMs remains a challenging problem due to the availability of the labeled data. Deep Convolutional Neural Networks (CNN) based fully supervised approaches have already been investigated and satisfactory classification performance have been obtained for the classification of WBM defect patterns. However, as they are fully supervised approaches, they require labeled data for training. Obtaining large amount of labeled data is a tedious and time consuming process. To overcome this, in this work we propose a CNN ensemble based semi-supervised approach, which make use of both labeled and unlabeled data for training. One of the main problem with CNN is that they often produce high-confident predictions, even for wrongly classified samples. We overcome this problem by the use of both Label-Smoothing and Ensembling. Comparative experiments on a large scale, public WBM dataset, WM-811K show that the proposed method is the new state-of-the-art, and we show that our approach outperforms other approaches even with relatively low amount of labeled data used for training.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.