Application of Deep Learning for Wafer Defect Classification in Semiconductor Manufacturing

Wafer map defect pattern classification is an important task for yield enhancement in semiconductor manufacturing. In this study, we proposed a framework to classify wafer map defect classification using a deep learning model combined with data augmentation techniques for handling the imbalance of the real-world semiconductor fabrication dataset WM-811 K. We compared the combination of the three scenarios for wafer map defect classification using deep learning model Convolutional Neural Network (CNN). The experimental results of our study show that both the two data augmentation methods using geometric transformation and conditional Generative Adversarial Networks (cGAN) improve the performance of the classifier based on CNN for wafer defect classification using different evaluation metrics. The geometric transformation provides better performance in classification compared to the cGAN in our study.