Multi-Layer Convolutional Neural Networks for Batik Image Classification

Abstract

Purpose: The purpose of this study is to enhance the classification of batik motifs through the implementation of a novel approach utilizing Multi-Layer Convolutional Neural Networks (CNN). Batik, a traditional Indonesian textile art form, boasts intricate motifs reflecting rich cultural heritage. However, the diverse designs often pose challenges in accurate classification. Leveraging advancements in deep learning, this research proposes a methodological framework employing Multi-Layer CNN to improve classification accuracy.

Methods: The methodology integrates Multi-Layer CNN architecture with an image dataset comprising various batik motifs, meticulously collected and preprocessed for uniformity. The CNN architecture incorporates convolutional layers of different sizes (3x3, 5x5, and 7x7) to extract unique features from batik images. Training options, including the Adam optimizer and validation frequency, are optimized based on parameters to enhance model efficiency and effectiveness.

Result: Results from the experimentation demonstrate significant improvements in classification accuracy, with an overall accuracy rate of 90.88%. Notably, precision and recall scores for individual batik motifs, such as Motif Cual Bangka and Motif Rumah Adat Belitung, reached remarkable levels, showcasing the efficacy of the proposed approach.

Novelty: This study contributes novelty through the integration of Multi-Layer CNN in batik classification, offering a robust and efficient method for identifying intricate batik motifs. Additionally, the research presents a pioneering application of deep learning techniques in preserving and promoting traditional cultural heritage, thereby bridging the gap between tradition and modern technology.