Dhruv Patel, a PhD student majoring in mechanical engineering, was awarded Honorable Mention for Best Student Paper at the SPIE Lithography Conference held on February 24-28 in San Jose, California. His paper, titled “Engineering Neural Networks for Improved Defect Detection, Classification, and Localization,” was co-authored with Viterbi Hughes Professor Assad Oberai and IBM researcher Ravi K. Bonam.
Their project sought to solve a major challenge in the semiconductor industry through the development of a deep-learning based workflow that could accurately detect, classify, and locate defects in semiconductors. By training a convolutional neural network, they were able to achieve high detection and classification accuracy with the ability to rapidly and accurately localize defects.
Below is a heat map used to visualize their predictions: