DOTA is a large-scale dataset for object detection in aerial images, containing 2806 images with a wide variety of object instances across 15 categories. Each image is approximately 4000×4000 pixels and includes objects of various scales, orientations, and shapes. The dataset is annotated by experts using oriented bounding boxes (OBBs) instead of axis-aligned ones, allowing for more accurate detection of objects in arbitrary orientations. The fully annotated dataset contains 188,282 instances, making it one of the largest and most diverse aerial object detection datasets. DOTA addresses the challenges of aerial object detection, including large variations in object size, arbitrary orientations, and dense object distributions. It also helps overcome dataset bias by reflecting real-world applications. The dataset is used to evaluate state-of-the-art object detection algorithms, demonstrating its effectiveness in representing real-world Earth Vision applications. DOTA provides a benchmark for object detection in aerial images and highlights the importance of using oriented bounding boxes for accurate detection. The dataset is also valuable for advancing research in computer vision and Earth Observation.DOTA is a large-scale dataset for object detection in aerial images, containing 2806 images with a wide variety of object instances across 15 categories. Each image is approximately 4000×4000 pixels and includes objects of various scales, orientations, and shapes. The dataset is annotated by experts using oriented bounding boxes (OBBs) instead of axis-aligned ones, allowing for more accurate detection of objects in arbitrary orientations. The fully annotated dataset contains 188,282 instances, making it one of the largest and most diverse aerial object detection datasets. DOTA addresses the challenges of aerial object detection, including large variations in object size, arbitrary orientations, and dense object distributions. It also helps overcome dataset bias by reflecting real-world applications. The dataset is used to evaluate state-of-the-art object detection algorithms, demonstrating its effectiveness in representing real-world Earth Vision applications. DOTA provides a benchmark for object detection in aerial images and highlights the importance of using oriented bounding boxes for accurate detection. The dataset is also valuable for advancing research in computer vision and Earth Observation.