Abstract
Front opening unified pods (FOUPs) are containers designed to hold wafers securely when they are carried. FOUPs thus need to be inspected on a timely basis before causing any loss for their failure. This work has developed a defect inspection program with the deep learning algorithm. The success of the program is demonstrated via inspection on four filters at the bottom of FOUPs. Filter images are collected from actual production lines are used to train the deep learning model. The performance of the model is even enhanced with a pre-trained VGG16 model as a feature extractor. The accuracy of the best model can reach 0.846 at the end.