UAV Detection and Classification using Deep Learning Techniques with YOLOv8

Accurately detecting and tracking drones in real-time poses main challenges due to factors such as varying scales, perspectives, occlusions, and environmental conditions. The proposed implementation helps in the identification of UAV objects in images and video footage using Deep Learning [DL] techniques, specifically the YOLOv8 architecture, by gathering a diverse dataset of images and video clips from the Roboflow 100 object detection benchmark dataset. Then the process is followed by pre-processing the data using resizing, normalization, and augmentation, which will be next initialized to the YOLOv8 architecture with pre-trained weights to estimate the performance using accuracy, precision, recall, and f1-score. The system finally involves integrating the classification system with a user-friendly interface for visualization of detection and classification in real time, where it results in which type of the drone is detected. The system achieved accuracy above 0.9 for all the classes of drones with different lights background and backgrounds in both images and video.