<![CDATA[Music is a universal form of art that has a significant impact on human life. In the digital era, managing increasingly large music collections requires an effective classification system to facilitate searching and storage. One of the growing methods is music genre classification, which helps organize music based on specific characteristics. This study explores the application of Convolutional Neural Network (CNN) and the MobileNet architecture for music genre classification based on spectrogram images. Spectrogram representation is used to convert audio signals into visual form, allowing the classification problem to be approached as an image classification task. The dataset used is GTZAN, consisting of six genres: blues, classical, country, hiphop, jazz, and metal. Image augmentation is applied to increase the diversity of training data, including rotation, translation, zooming, brightness adjustment, and horizontal flipping. The evaluation results show that the CNN-MobileNet model achieves an overall accuracy of 83%, with a macro precision of 85%, macro recall of 83%, and macro F1-score of 84%. The classical genre achieved the best performance with an F1-score of 93%. This research demonstrates that spectrogram-based music genre classification using CNN-MobileNet is an effective approach for automatic music recognition tasks]]>
Copyrights © 2025
