Recent growth in deep learning and computer vision has opened up many opportunities for advanced intelligent systems. While the dataset quality plays a crucial role in the training phase and also affects the performance of a model, creating a reliable and diverse dataset appears to be challenging. Therefore, data augmentation can be used as a preprocessing step with a view to tackling the problem. In this paper, we conduct a comprehensive analysis on different augmentation strategies to investigate their impact on vehicle detection in top-down videos captured by drones. Our experiments show that by adding similar data, randomly rotating and cropping images with respect to the model's input size, we have remarkably increased the accuracy of YOLOv3, one of the state-of-the-art and real-time object detection methods.