Deep learning algorithms for batik design classification

Lee, Xiao Xuan (2024) Deep learning algorithms for batik design classification. Project Report. Melaka, Malaysia, Universiti Teknikal Malaysia Melaka. (Submitted)

Text (Full Text) Deep learning algorithms for batik design classification.pdf - Submitted Version Download (2MB)

Abstract

Batik is widely known as one of the unique identifiers among Southeast Asia countries including Malaysia. Batik industry in Malaysia holds an important place in the craft-based industry. However, the batik industry in Malaysia lack of a comprehensive and standardized dataset which is crucial for developing an automated classification system. Besides, the traditional method of batik classification needs trained experts to visually inspect and analyse the intricate pattern. Therefore, this study indicates to compile and construct a new dataset of Malaysia batik for image classification. This study also aims to develop a batik classification system using deep learning algorithms that classify between Indonesia and Malaysia batik. A dataset with 1825 images including 949 images for Indonesia batik and 876 images for Malaysia batik is utilised. CNN models including MobileNet v2, YOLO-v8 and LeNet-5 is selected and the performance of these three CNN models is further investigated based on their accuracy, loss and confusion matrix. As a result, all three models reach a high accuracy which is 97.79% for MobileNet v2, 98.80% for YOLO-v8 and 92.94% for LeNet-5.

Actions (login required)

View Item