DOTA: A Large-scale Dataset for Object Detection in Aerial Images

DOTA是武汉大学研究团队创建的大规模航空图像目标检测数据集，旨在推动遥感图像中目标检测技术的研究。该数据集包含从不同传感器和平台收集的高分辨率航拍图像，涵盖飞机、船舶、车辆、桥梁等18个常见类别，采用任意四边形标注方式。DOTA数据集已成为遥感领域最经典的目标检测基准之一，广泛应用于旋转目标检测算法的开发与评估。DOTA is a large-scale aerial image object detection dataset developed by the research team from Wuhan University, aiming to advance the research on object detection technology in remote sensing images. This dataset includes high-resolution aerial images collected from various sensors and platforms, covering 18 common categories such as aircraft, ships, vehicles, bridges and other objects, and adopting arbitrary quadrilateral annotation method. The DOTA dataset has become one of the most classic object detection benchmarks in the remote sensing field, and is widely used for the development and evaluation of rotated object detection algorithms.

搜集汇总

背景与挑战

背景概述

DOTA是一个大规模航空图像目标检测数据集，包含多个版本（如v1.0、v1.5、v2.0），图像尺寸从800x800到20,000x20,000像素，标注为任意四边形，覆盖15-18个物体类别，实例数量高达数百万。该数据集专为开发和评估航空图像中的目标检测算法设计，具有尺度、方向和形状多样性，并通过挑战赛和持续更新推动领域发展。