Enhancing YOLO with Adversarial and Transfer Learning for UAV-Based Urban Wildlife Conservation

Urban wildlife conservation has gained more importance worldwide, particularly in fast urbanizing regions like Kingdom of Saudi Arabia (KSA), where habitat fragmentation and biodiversity loss are major concerns. Unmanned aerial vehicle (UAV) imagery has emerged as a promising solution for wildlife monitoring, providing high-resolution, real-time data collection even in difficult-to-access urban areas. Nonetheless, the effective utilization of UAV images is seriously hampered by the metropolitan settings and interference from human activity. While deep learning (DL) models provide effective solutions for wildlife monitoring in UAV imagery, the scarcity of high-quality training datasets often limit their effectiveness. To the best of our knowledge, this is the first research to resolve these issues harnessing the combined power of adversarial and transfer learning. The research enhances the state-of-the-art YOLOv8 (You Only Look Once, version 8) model by integrating Generative Adversarial Networks (GANs) for data augmentation (DA) and EfficientNet-B3 for advanced feature extraction. Two research datasets namely, the Wildlife Animals Image Dataset (WAID) and the Animal Images Detection (AID) are used to evaluate the proposed model. Three comprehensive experimental analyses—DA, ablation, and cross-dataset validation (CDV) are carried out to prove its efficacy in urban wildlife monitoring. The results highlight the potential of the proposed model to considerably enhance detection accuracy and contribute to sustainable urban wildlife conservation efforts, which are consistent with the United Nations Sustainable Development Goals (UN SDGs).