Military Aircraft Identification Using VGG-16: A Deep Learning Approach to Multi-Class Classification

This study investigates the utility of the VGG-16 CNN model for identifying military aircraft. For training and testing purposes, a dataset of 2,500 labeled images was used, classified as Fighter, Bomber, UAV, and Reconnaissance classes. Preprocessing involved size reduction of images to 256×256 pixels, normalization, and data augmentation techniques such as horizontal flipping, rotations, and zooming. Validation accuracy results for fine-tuned VGG-16 was 93.14%. The model designed here, including convolutional and pooling layers along with batch normalization while being optimized using the Adam algorithm, showed high robustness in extracting unique features of aircraft. Parameters like precision, recall, and F1-score could be employed for evaluating this model that has shown to minimize misclassification rates. Therefore, results present it as a very promising approach toward real-time applications in military surveillance and defense systems. Future work will include increasing the dataset, model optimization, and deployment to better and realistic environments to improve adaptability and accuracy.