Browsing by Author "Jung, S. Y."

Now showing 1 - 2 of 2

Deep Learning Based AOI System with Equivalent Convolutional Layers Transformed from Fully Connected Layers
(2019-01-01)
Tsai, Y. H.
;
Lyu, N. Y.
;
Jung, S. Y.
;
Chang, K. H.
;
Chang, J. Y.
The rise of deep learning, especially in the realm of computer vision, paves ways of leveraging automatic optical inspection systems to a higher level. Convolutional neural networks and its derivatives might be the most widely used architectures for defect inspection tasks. In real cases the amount of collected data is often not large, so transferring learning and data augmentation are necessary. In this paper, we explain some details how we implement the deep learning based AOI system where fully connected layers are replaced by convolutional layers, then a classification heat map is output after post-processing. We examine the performance of our model with two data sets collected in industrial manufacturing cases. We further propose an idea to transfer models pretrained on augmented data of different sizes cropped from original image to the present classification task for possible improvements of the performance.
Defect Detection on Randomly Textured Surfaces by Convolutional Neural Networks
(2018-01-01)
Jung, S. Y.
;
Tsai, Y. H.
;
Chiu, W. Y.
;
Hu, J. S.
;
Sun, C. T.
Automatically detecting the defects on the randomly textured surfaces for industrial purpose is a demanding procedure due to the ambiguity between defects and textures, lack of defect-labeled data and the must-have extreme accuracy. In this paper we proposed a procedure as the beginning of automating the defect detection on woods with randomly textured surfaces by employing 3 different architectures of convolutional neural networks. The deep convolutional neural network resulted in 99.80% accuracy, discriminating among normal wood and the other 4 types of defects images. The models were evaluated and understood by visualizing the saliency maps. The results from our work implies that other industrial images with defects on randomly textured surfaces may apply the similar procedures to accelerate the automating of defect detection and progressing of industry 4.0.