Utilizing convolutional neural networks for the classification and preservation of Kalinga textile patterns

Abstract

This study introduces a system that utilizes Convolutional Neural Networks (CNN) to categorize Kalinga textiles in a structured manner. The main objective is to systematically identify and name the patterns found in these textiles. The research uses a dataset that includes ten different categories of Kalinga textiles. Metrics such as accuracy, precision, recall, and F1 Score are used to assess the performance of the system. The outcomes demonstrate high precision values between 0.8 and 1.00, showcasing the model's proficiency in precisely classifying and labeling the patterns of Kalinga textiles. Similarly, the recall values, which vary from 0.75 to 1.00 for each category, underscore the model's effectiveness in categorization. These results highlight the system's capability to recognize and categorize Kalinga textiles, with recall values providing strong evidence of its reliability. F1 scores, which consider both precision and recall, range from 0.86 to 0.97 across the categories, indicating the model's accuracy in classification. The introduced technique for image identification shows promise for identifying and categorizing Kalinga textiles, thereby contributing to the preservation and promotion of this cultural heritage. It offers a valuable tool for researchers, enthusiasts, and cultural institutions. Future research could focus on expanding the dataset to improve the model's robustness and exploring its application to other areas of textiles. Continuous enhancements to the model, based on user feedback and technological advancements, will ensure its ongoing effectiveness and relevance.