Memory Enhanced Dynamic Conditional Random Fields Embedded Pairwise Potential CNN for Fabric Defects Identification

Memory Enhanced Dynamic Conditional Random Fields Embedded Pairwise Potential CNN for Fabric Defects Identification

  IJETT-book-cover           
  
© 2021 by IJETT Journal
Volume-69 Issue-10
Year of Publication : 2021
Authors : B. Vinothini, S. Sheeja
DOI :  10.14445/22315381/IJETT-V69I10P229

How to Cite?

B. Vinothini, S. Sheeja, "Memory Enhanced Dynamic Conditional Random Fields Embedded Pairwise Potential CNN for Fabric Defects Identification," International Journal of Engineering Trends and Technology, vol. 69, no. 10, pp. 227-234, 2021. Crossref, https://doi.org/10.14445/22315381/IJETT-V69I10P229

Abstract
In the textile industry, defect identification is the most crucial process for localizing Fabric Defects (FDs) and enhancing yarn quality. In earlier centuries, many techniques were discussed to identify the FDs automatically. Among those, a hybrid technique called Pairwise-Potential Activation Layer in Convolutional Neural Network (PPAL-CNN) localizes the fine structures in textile imagery through integrating dynamic AL on CNN and a PP factor in the Conditional Random Fields (CRFs). However, this CRF should be supplied a priori rather than learned. It was hard for a complex interaction between FDlabels/classes while performing multiple, or long-range dependencies exist. Therefore, this paper proposes an Enhanced PPAL-CNN (EPPAL-CNN) technique to manage the complex structure interaction of FDs. First, the CRF is extended by integrating external memory strategies stimulated from the memory networks and thus facilitating CRFs for interpretation beyond localized characteristics and have access to the complete image. It encompasses the memory and Dynamic CRF (DCRF) layers. The memory layer is partitioned into input, output, and current input memory. The interpretations of input and output memory have interacted through an attention model in which weights are calculated by the relation of an input and a current input memory. Then, an outcome of the memory layer is taken as input to the DCRF layer. The DCRFs are simplified linear CRFs and are used for defining the shared hidden state and complicated relation between labels. Its factorial construction includes the relations among cotemporally labels, explicitly modeling constrained likelihood dependencies among various labels. So, a high-level Markov dependency among labels is modeled by considering the external memory. Finally, the investigational outcomes exhibit that the EPPAL-CNN achieves 93.36% accuracy compared to the PPAL-CNN technique using the TILDA database.

Keywords
Fabric defects, Defect identification, CNN, Pairwise-potential activation, CRFs, Memory network.

Reference
[1] Ngan, H. Y., Pang, G. K., & Yung, N. H. (2011). Automated fabric defect detection—a review. Image and Vision Computing, 29(7) 442-458.
[2] Tsang, C. S., Ngan, H. Y., & Pang, G. K.. Fabric inspection based on the Elo rating method. Pattern Recognition, 51 (2016) 378-394.
[3] Long, J., Shelhamer, E., & Darrell, T. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2015) 3431-3440.
[4] Ronneberger, O., Fischer, P., &Brox, T. U-net: convolutional networks for biomedical image segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention, Springer, Cham, (2015) 234-241.
[5] Badrinarayanan, V., Kendall, A., & Cipolla, R. SegNet: a deep convolutional encoder-decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(12) (2017) 2481-2495.
[6] Zheng, S., Jayasumana, S., Romera-Paredes, B., Vineet, V., Su, Z., Du, D., ... & Torr, P. H. Conditional random fields as recurrent neural networks. In Proceedings of the IEEE International Conference on Computer Vision, (2015) 1529-1537.
[7] Çelik, H., Dülger, L. C., Topalbekiroglu, M., & Rosa, J. L. G. Application of neural networks (NNs) for fabric defect classification. In Artificial Neural Networks Models and Applications. InTech(2016).
[8] Ouyang, W., Xu, B., Hou, J., & Yuan, X. Fabric defect detection using activation layer embedded convolutional neural network. IEEE Access, 7(2019)70130-70140.
[9] Liu, Z., Li, C., Zhao, Q., Liao, L., & Dong, Y. A fabric defect detection algorithm via context-based local texture saliency analysis. International Journal of Clothing Science and Technology, 27(5) (2015) 738-750.
[10] Li, Y., Zhao, W., & Pan, J. Deformable patterned fabric defect detection with fisher criterion-based deep learning. IEEE Transactions on Automation Science and Engineering, 14(2) (2017) 1256-1264.
[11] Jia, L., Zhang, J., Chen, S., & Hou, Z. Fabric defect inspection based on lattice segmentation and lattice templates. Journal of the Franklin Institute, 355(15) (2018) 7764-7798.
[12] Yapi, D., Allili, M. S., &Baaziz, N. Automatic fabric defect detection using learning-based local textural distributions in the contourlet domain. IEEE Transactions on Automation Science and Engineering, 15(3) (2018) 1014-1026.
[13] Mei, S., Wang, Y., & Wen, G. Automatic fabric defect detection with a multi-scale convolutional denoising autoencoder network model. Sensors, 18(4) (2018) 1064.
[14] Jeyaraj, P. R., & Samuel Nadar, E. R. Computer vision for automatic detection and classification of fabric defects employing deep learning algorithm. International Journal of Clothing Science and Technology, 31(4) (2019) 510-521.
[15] Lin, S., He, Z., & Sun, L. Defect enhancement generative adversarial network for enlarging data set of microcrack defect. IEEE Access, 7, (2019) 148413-148423.
[16] Hu, G., Huang, J., Wang, Q., Li, J., Xu, Z., & Huang, X. Unsupervised fabric defect detection based on a deep convolutional generative adversarial network. Textile Research Journal, 90(3-4) (2020) 247-270.
[17] Jing, J., Wang, Z., Rätsch, M., & Zhang, H. Mobile-Unet: an efficient convolutional neural network for fabric defect detection. Textile Research Journal, (2020) 1-13.