Convolutional neural network-based mode selection and defect classification for image synthesis

摘要

Systems and methods are disclosed for classifying defects using hot scans and convolutional neural networks (CNN). The primary scanning modes are identified by the processor and a hot scan of the wafer is performed. Interest defects and Newson's data are selected and the images of these areas are captured using one or more secondary scanning modes. The image sets are collected and divided into subsets. CNN is trained using image subsets. An ideal secondary scanning mode is determined and a final hot scan is performed. The defects are filtered and sorted according to the final hot scan and the ideal secondary scanning mode. The disclosed systems for classifying defects use scanning electron microscopes as well as image data acquisition subsystems such as processors and electronic databases.