Klasifikasi Senjata Adat Indonesia Berbasis Deep Learning dengan Model InceptionV3

This study develops a traditional Indonesian weapon image classification system by expanding the dataset to include nine categories, totaling 1,385 images. The model is built using the InceptionV3 Convolutional Neural Network (CNN) architecture with a transfer learning approach. The dataset is split into 70% training data, 20% validation data, and 10% test data. All images are resized to 299×299 pixels and normalized to match the model's input requirements. Several training scenarios were evaluated using different optimizers (Adam, RMSprop, SGD, Adamax) and data augmentation techniques. The best results were achieved using a combination of Adam and RMSprop optimizers with data augmentation, reaching an accuracy of 96% and a loss value as low as 0.21. Data augmentation proved effective in enhancing data variability, reducing overfitting, and improving the model’s generalization capabilities. The findings demonstrate the effectiveness of the InceptionV3 architecture for classifying traditional weapon images and contribute to cultural preservation through the development of an accurate and efficient automatic classification system.