<![CDATA[The digital preservation of cultural heritage has become increasingly important in sustaining local traditions amid rapid modernization. Balinese traditional kites represent a distinctive form of intangible cultural heritage with unique visual characteristics; however, their identification and classification are still largely based on subjective expertise. This research develops a Convolutional Neural Network (CNN)-based model for image classification to automatically recognize three primary types of Balinese traditional kites: Bebean, Janggan, and Pecukan. Beyond technical implementation, this research contributes to the development of a culturally specific visual dataset, addressing the limited representation of local heritage objects in mainstream computer vision research, which is predominantly based on global datasets of generic objects. A balanced dataset of 2,400 images was constructed and evaluated using 5-Fold Cross Validation to assess model stability and generalization capability. The proposed CNN model achieved an average validation accuracy of 91.5%, with balanced precision, recall, and F1-score across folds. Further evaluation on an independent test set of 282 images resulted in an accuracy of 87.94%, indicating a generalization gap of approximately 4%, which remains within an acceptable range. The results demonstrate that CNN-based classification can effectively support structured digital documentation of traditional kites. This study highlights the potential of computer vision not only as a technical tool, but also as a strategic approach to advancing data-driven cultural preservation and expanding AI applications within localized cultural contexts.]]>
Copyrights © 2026
