Data Augmentation Method on Drone Object Detection with YOLOv5 Algorithm

This study aims to evaluate and compare various types of data augmentation in drone object detection using the YOLOv5 algorithm. Data augmentation is an important technique in image data processing to increase the diversity and quality of training data, so as to improve the performance of object detection models. In this study, several different types of data augmentation were applied to the existing drone object detection dataset. This research method consists of fourteen types of data augmentation, namely Flip, Rotate, Crop, Rotation, Shear, Hue, Blur, Grayscale, Saturation, Brightness, Exposure, Noise, Cutout, and Mosaic. The use of the Mosaic method as the newest type of data augmentation is expected to provide better results in drone object detection. The test is carried out by measuring the Precision-Recall evaluation metrics for each type of data augmentation. The results showed that the Mosaic augmentation type gave the highest Precision-Recall value of 0.993 compared to other augmentation types. The use of Mosaic as a data augmentation in drone object detection with the YOLOv5 algorithm can significantly improve object detection performance.