Classification of Solo Batik patterns using deep learning convolutional neural networks algorithm

Abstract

The ideology of the Solo Batik pattern has not been conveyed to the public. In addition, a lot of people are unaware that batik contains particular patterns that are also used for particular activities. This study uses a convolutional neural network model to categorize 9 different Solo Batik patterns according to their use of elaborate geometric shapes, complicated symbols, patterns, dots, and natural designs. With 1 to 4 hidden layers, we aim to select the number of hidden layers that yields the highest accuracy. A 100×100 pixel image is used as the input. The feature extraction process then makes use of 3×3 feature maps from three convolution layers. The dropout regularization is then added, with settings ranging from 0.1 to 0.9. The Adam algorithm is also used in this model to perform optimization. The 3-layered convolutional neural networks (CNN) with a dropout value of 0.2, run in 20 epochs, produced accuracy results of 97.77%, which was the highest. Additionally, it can be inferred that applying a certain number of hidden layers and adding right dropout regularization values has an impact on raising the accuracy score.