This dataset has 32,000 remote sensing images in UAV scenes of tiny objects with labels. The images in the dataset contain seven types of ground vehicle targets, and all of them are tiny targets. Data augmentation is one of the simplest and most effective methods to enhance the performance of tiny object detection. The dataset was built based on the Mixup data enhancement algorithm, obtained from the original 16,000 labeled ground vehicle tiny targets in UAV scenes by data enhancement.Compared with many publicly available datasets, this dataset has the following advantages: large dataset, small target size, complex distribution, etc. It can be widely used in the field of tiny object detection.
The dataset contains two folders. One folder is "JPEGImages", which contains 32,000 images of ground vehicle targets in UAV scenes. The other folder is "Annotations", which contains label data for seven types of vehicle targets. "JPEGImages" corresponds to the data in "Annotations".