Comparison of Classification of Songket Fabric Types Using AlexNet and VGG19 (Visual Geometry Group) Method

This study aims to evaluate and compare the performance between deep learning models AlexNet and VGG19 in Songket fabric classification. Due to its complex patterns and subtle differences, Songket classification must be accurate. The datasets in this study are various types of Songket images and all datasets are classified by type for easy analysis. After intensive learning and evaluation, VGG19 is a superior classifier than AlexNet. The highest performance is achieved by the VGG19 method in terms of performance measure accuracy, precision, recall, and F1 score, which may be due to the increase in depth and better extraction of some detailed visual features from complex images. Although these results have substantial practical implications, some issues need to be further discussed before optimizing the results. Hyperparameters, such as learning rate or batch size, can be changed to optimize the speed and accuracy of the model. In addition, the diversity of the data should be increased by using data augmentation techniques to ensure that the model generalization to market conditions is possible. More complex additions (lighting changes, texture distortion simulation, or others) can also contribute to improving the robustness of the trained model to these disturbances. The conclusion of the research is the importance of improving the accuracy and usefulness of single fabric classification. This will result in its application in heritage preservation and textile development.