Klasifikasi Motif Citra Batik Menggunakan Convolutional Neural Network Berdasarkan K-means Clustering

Abstract

Batik has several motifs and patterns so it is necessary to identify certain objects in an image, one of which is the recognition of the image of Yogyakarta batik using the Convolutional Neural Network (CNN) method which is already popular in the use of image data classification. The introduction of batik imagery aims to contribute to the digitization of batik image data and at the same time provide information on types of batik to the public. The batik image recognition process using CNN in this study combines the image segmentation process and the enhancement process with median filters and sharpening. The segmentation process carried out before CNN aims to help separate foreground objects from objects that are not needed in the background. The segmentation process that is commonly used is using K-means Clustering. Where K-means Clustering is used to group data in the same category. Furthermore, the enhancement process using the median filter and sharpening was carried out separately to compare the batik image classification process using CNN based on K-means Clustering from the median filter results and the sharpening results. The batik image classification process with CNN based on K-means Clustering on the median filter resulted in an accuracy value of 100%. Meanwhile, the batik image classification process with CNN based on K-means Clustering from the sharpening results resulted in an accuracy value of 80%.

Convolutional Neural Networks for Classification Motives and the Effect of Image Dimensions

“…The batik pictures were categorized using a Convolutional Neural Network (CNN), with K-means clustering as the underlying method. Clustering was used to obtain preprocessed results from the median filter of [13].…”