<![CDATA[This research is motivated by the declining ability of the younger generation to recognize the physical forms of traditional musical instruments from East Nusa Tenggara (NTT). This study aims to apply a Convolutional Neural Network (CNN)- based Deep Learning algorithm to classify seven types of musical instruments: Sasando, Moko, Gong, Gendang, Foydoa, Jungga, and Knobe Oh. A dataset of 610 images was used, split 80% for training and 20% for testing, and subjected to preprocessing steps, including resizing and data augmentation. The performance of a Custom CNN architecture was compared against VGG19 (fine-tuning) across variations of learning rates (0.001, 0.0001, and 0.00001). The results showed that the Custom CNN achieved optimal performance at a learning rate of 0.0001, with a training accuracy of 91.63% and a validation accuracy of 82.50%. Meanwhile, the fine-tuned VGG19 model achieved 100% training accuracy and 93.33% validation accuracy. Confusion Matrix evaluation on the test data demonstrated that the best model achieved 100% accuracy, indicating that the system successfully extracts and recognizes the visual features of each instrument.]]>
Copyrights © 2026
