FINE-GRAINED CLASSIFICATION OF MILITARY AIRCRAFT USING PRE-TRAINED DEEP LEARNING MODELS AND YOLO11

This research examines the potential of pre-trained deep learning models for the fine-grained classification of military aircraft, to achieve accurate identification and extraction of unique tail numbers. The study uses a publicly available dataset comprising 43 classes of military aircraft, with a total of 24,164 images for training and 6,042 images for testing. The performance of five distinct pre-trained convolutional neural network (CNN) architectures, including DenseNet121, MobileNetV2, ResNet50, ResNet101, and VGG19, is evaluated and compared. Further more, the paper examines the effectiveness of the YOLO11 model family for aircraft classification, particularly emphasizing the YOLO11x-cls model’s superior performance. The study analyses the training results and confusion matrix of the YOLO11x-cls model, demonstrating its accuracy and ability to generalize well to unseen data. This work contributes to the advancement of AI-powered image recognition for military aviation applications, potentially improving data collection, monitoring, and analysis processes.