Image-based fabric retrieval technique can help to develop new fabrics and manage products. Efficiently extracting features from fabric images is the key to enhance the practicality of this technology. In this paper, convolutional neural network is trained with a dataset of 19,894 different yarn-dyed fabric patterns. Center loss architecture is added to further improve the discriminative power of the network. By properly sampling from original images, the network model can efficiently extract discriminative features and achieve a retrieval accuracy of 99.89% on our test set. This performance maintains well when simpler deep architecture is used, but decreases quickly if the contents of fed fabric image are reduced.
CITATION STYLE
Wang, X., Wu, G., & Zhong, Y. (2019). Fabric identification using convolutional neural network. In Advances in Intelligent Systems and Computing (Vol. 849, pp. 93–100). Springer Verlag. https://doi.org/10.1007/978-3-319-99695-0_12
Mendeley helps you to discover research relevant for your work.