Mamluatul Hani'ah, Rendy Septian Pradana, Muhammad Afif Hendrawan, Vivi Nur Wijayaningrum, Vit Zuraida, Moch. Zawaruddin Abdullah, Mungki Astiningrum
Batik, a traditional textile art recognized by UNESCO as an Intangible Cultural Heritage, holds profound cultural and aesthetic values across Indonesia. To support cultural preservation through digital recognition, this study proposes a classification approach for Malang batik motifs using a parallel Convolutional Neural Network (CNN) architecture. A dataset of 622 images across 31 motif classes was collected directly from three Malang batik production centers and enriched using various augmentation techniques to overcome data scarcity. The proposed model employs two parallel convolutional branches with different kernel sizes to capture multi-scale visual features, followed by concatenation, global average pooling, and fully connected layers for final classification. Model training was performed using a 5-fold cross-validation scheme with the Adam optimizer and ELU activation function. The proposed parallel CNN architecture was designed to extract richer and more diverse feature representations, improving classification accuracy compared to conventional sequential CNN models. Experimental results demonstrated that augmentation significantly influenced performance, where geometric transformations such as rotation achieved stable accuracy above 90%. The highest performance was obtained using a combination of all augmentations, achieving 99.59% accuracy, 99.60% precision, 99.50% recall, and 99.50% F-measure. However, this improvement came with higher computational costs, as runtime and memory consumption were substantially increased compared to simpler augmentation strategies. These findings highlight a clear trade-off between accuracy and computational efficiency. Overall, the proposed method demonstrates the potential of parallel CNNs and augmentation in enhancing batik motif recognition. © 2025 IEEE.
Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia