The rapid deployment of drones in commercial and recreational applications has increased concerns regarding the exploitation of drones in unlawful ways. AI-based autonomous drone detection systems are one of the potential solutions that have been proposed to overcome the issues related to drone misuse such as transferring illegal material, spying in restricted areas, violating privacy of people, etc. Furthermore, vision based autonomous drone detection can also be used for the purpose of localization in UAV swarm systems. Detection of drones is a quite difficult task as the system has to identify drones among similar kinds of objects like birds, airplanes, etc. In addition to this, drone detection models need to be trained using a significant amount of data to obtain accurate and precise predictions. However, in the case of real-time detection, a highly configured computational device (GPU) is required. The idea behind this paper is to overcome these limitations of existing detection algorithms by presenting the model based on a single-shot object detector called You Only Look Once version 8 (YOLOv8) which can be trained using transfer learning and data augmentation to deal with the data shortage. The predictions of the trained model are validated using precision, recall, and mAP (Mean Average Precision) evaluation measures. The presented model achieved a 91.1% mAP, and a 2.36% improvement over the model based on the older YOLO version (YOLOv7) and trained on the same dataset.