Thai weaving pattern classification using convolutional neural networks

Abstract

The creation of Thai weaving patterns based on reproduction of original patterns with adapted patterns inspired by designer is a way to reserve Thai textiles. Although weaving pattern designs are useful for promoting Thai textiles such as in casual attires, preserving original Thai patterns is still needed. This work aimed to classify the images of original Thai ethnic and adapted weaving patterns. In this paper, we trained 28 original Thai ethnic patterns using the convolutional neural network. We cropped and preprocessed the weaving images to a binary format. The data augmentation method was also used to increase the number of weaving patterns for training the convolutional neural network. The model was tested in the real world by using test patterns from Google images and gave the results of 0.90, 0.92, and 0.90 for precision, recall, and F1 score, respectively.