UAV Detection, and Explanation Using a Tuned Transfer Learning Model

Computer vision is one of computer science's most important subfields. The rapid evolution of the camera, video, and image applications led to this significance. In the past years, drones have been rapidly developed. Most of the time, they have been employed for privacy violations, espionage, and military attacks in the past several years. The rotary drone is one of the different Unmanned Aircraft vehicles (UAV) used for various reasons. This paper uses a deep learning-based model to detect and classify rotary drones on two datasets. The first image dataset shows the drones with a white or transparent background, while the remaining images include a different background without a drone gathered from the second dataset. The second dataset consists of images with diverse backgrounds, some of which contain drones and others without being used for testing the system. The suggested system includes three stages. The first is preprocessing, such as Normalized and resized images. The second parts represent the tuned VGG16 model that trained for twenty epochs passed since the beginning of the training process to produce results. Finally, the explanation stage is used to ensure the correctness of decision-making. The results indicate an efficient drone classification with 99% accuracy.