Textile pattern style classification based on popular mixture enhancement and attribute clustering

Intending to the issue that traditional textile pattern classification methods have insufficient training samples and ignore the attributes possessed by the style objects, this article designs a textile pattern style classification method relied on popular mixture enhancement and attribute clustering. Firstly, the entropy discretisation technique is introduced to optimise the image attribute clustering method, and discrete values are used to represent the discretised data to eliminate the metric differences. Secondly, the original textile images are popularly mixed and enhanced according to the mixing parameter. And the visual feature intersection of the enhanced pattern is used as an object mask by using two-channel CNN output to map onto the original image to obtain an object-level image, and the features are enhanced by the channel attention mechanism. The simulation results show that the accuracy and average precision of the proposed method have a mean value of 83.59% and 91.36%, respectively. Copyright © The Author(s) 2024.