Mixup-based Neural Network for Image Restoration and Structure Prediction from SEM Images

• Published in: IEEE transactions on instrumentation and measurement 2024-01, Vol.73, p.1-1
• Main Authors: Park, Junho, Cho, Yubin, Hwang, Yeieun, Ma, Ami, Kim, QHwan, Chang, Kyu-baik, Jeong, Jaehoon, Kang, Suk-Ju
• Format: Article
• Language: English
• Subjects: Algorithms; classification; Convolutional neural networks; Datasets; Deep learning; Image analysis; image denoising; Image restoration; Machine learning; regression; Scanning electron microscope (SEM); Scanning electron microscopy; super resolution; Superresolution; Task analysis; Training; Transformers
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2024.3366569

Scanning electron microscopy (SEM) has been widely used for the semiconductor industry since it provides high-resolution details of the semiconductor. However, there is a gap in research for various tasks (i.e., image restoration and structure prediction) in SEM datasets collected under various conditions. Therefore, we introduce a new SEM dataset with diverse characteristics such as energy, noise, current with various levels for image restoration and structure prediction. Furthermore, we propose a new deep learning-based method for this dataset. The method consists of two stages: image restoration stage and structure prediction stage. In the image restoration stage, we design the transformer-based architecture to utilize pixel information widely. In the structure prediction stage, we introduce a novel training algorithm, SEMixup, and a novel CNN-based network, SEM-SPNet. Specifically, SEMixup overcome the generalization and robustness of SEM-SPNet by implicitly interpolating a pair of samples and their labels. Experiments demonstrate that our method achieves state-of-the-art results across all dataset conditions. This work expands the possibilities of SEM image analysis using deep learning, contributing to the semiconductor industry.