2018 · 20 citations · 8 references
ComputingComputer EngineeringComputer ScienceArtificial IntelligenceComputer VisionMachine LearningDeep LearningDefect ClassificationDefect DetectionImage ClassificationIndustrial InspectionNeural NetworksAutomatic ClassificationConvolutional Neural NetworksConvolutional Neural Networks For InspectionDeep Learning For InspectionDeep Learning-based ClassificationImage-based Defect ClassificationThrough-silicon Via DesignThrough-silicon Vias
Deep Neural Network technology has shown impressive performance in visual recognition problems such as defect image classification that depends on the skills and experiences of individual inspectors. We selected a Through-Silicon Via (TSV) process which has relatively few defect types to adapt deep networks as the first test bed. In this paper, we propose Convolutional Neural Network (CNN)-based defect image classification model derived from Residual Network which ranked first in image classification competitions such as ILSVRC and COCO 2015 with 4.62% test error. However, merely bringing the well-known architecture to the defect classification task was unable to resolve dataset problems: imbalance, ambiguity and inconsistency. We maximized the classification performance to 97.1% accuracy on the long-term dataset by optimizing classifier and cleansing the dataset. Our model can lessen defect classification work done by human by as much as 78.6%.
8
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren et al. · 2016 · 222.1K citations · Full text
ImageNet: A large-scale hierarchical image database
Jia Deng, Wei Dong, Richard Socher et al. · 2009 IEEE Conference on Computer Vision and Pattern Recognition · 2009 · 61.8K citations
Gradient-based learning applied to document recognition
Yann LeCun, Léon Bottou, Yoshua Bengio et al. · Proceedings of the IEEE · 1998 · 58.2K citations · Full text