A deep learning-based approach for defect classification with context information in semiconductor manufacturing

This thesis presents some methodological and experimental contributions to a deep learning-based approach for the automatic classifi cation of microscopic defects in silicon wafers with context information. Canonical image classifi cation approaches have the limitation of utilizing only the information contained in the images. This work overcomes this limitation by using some context information about the defects to improve the current automatic classifi cation system.