DOTA：用于航拍图像目标检测的大规模数据集

DOTA 数据集便是用于航拍图像中目标检测的图像数据集，其被用于发现和评估航拍图像中的物体。 这些图像来源包含不同传感器和平台，包括 Google Earth、JL–1 卫星拍摄，以及中国资源卫星数据和应用中心的 GF–2 卫星拍摄。 数据集共计 2806 幅航拍图，每张图像的像素尺寸在 800*800 到 4000*4000 的范围内，其中包含不同尺度、方向和形状的物体。之后这些 DOTA 图像经由专家使用 15 个常见目标类别进行注释，包括：飞机，轮船，储罐，棒球场，网球场，篮球场，地面跑道，港口，桥梁，大型车辆，小型车辆，直升机，环形交叉路口，足球场和篮球场。 完全注释的 DOTA 图像包含 188282 个实例，每个实例均由任意四边形进行标记。 数据集中用标注有定向的边界框标注实例 以下是该数据集的详细信息： DOTA 数据集 发布机构： 武汉大学 包含数量：2806 张航拍图片 数据格式：txt.pnd 图片尺寸：800 * 800~4000*4000 数据大小：35 GB 类别数：15 类 If you have any the problem or feedback in using DOTA, please contact Gui-Song Xia at [e-mail], Xiang Bai at [e-mail], Jian Ding at [e-mail], Zhen Zhu at [e-mail]. Zhipeng Lin at [e-mail],

The DOTA dataset is an image dataset designed for object detection in aerial imagery, which is utilized to discover and evaluate objects within aerial photographs. These images are sourced from various sensors and platforms, including Google Earth, JL-1 satellite imagery, as well as data from the GF-2 satellite operated by the China Resources Satellite Data and Application Center.

The dataset contains a total of 2806 aerial images, with each image having a pixel dimension ranging from 800×800 to 4000×4000, encompassing objects of different scales, orientations and shapes. Afterwards, these DOTA images were annotated by experts using 15 common object categories, including: airplane, ship, storage tank, baseball diamond, tennis court, basketball court, ground runway, harbor, bridge, large vehicle, small vehicle, helicopter, roundabout, soccer field and basketball court. Fully annotated DOTA images contain 188,282 instances, each marked with an arbitrary quadrilateral. Instances in the dataset are annotated with oriented bounding boxes.

Below are the detailed specifications of the DOTA dataset:
- Issuing institution: Wuhan University
- Total number: 2806 aerial images
- Data formats: .txt, .pnd
- Image size: 800×800 ~ 4000×4000
- Data size: 35 GB
- Number of categories: 15

If you have any problems or feedback when using DOTA, please contact the following authors: Gui-Song Xia at [e-mail], Xiang Bai at [e-mail], Jian Ding at [e-mail], Zhen Zhu at [e-mail], Zhipeng Lin at [e-mail].

搜集汇总

背景与挑战

背景概述

DOTA是一个专为航拍图像目标检测设计的大规模数据集，包含2806幅来自不同传感器（如Google Earth和卫星）的图像，尺寸范围从800*800到4000*4000像素，覆盖15个常见目标类别，如飞机、轮船和车辆，共计188282个实例，使用任意四边形进行精确标注。该数据集由武汉大学发布，数据大小为35 GB，适用于评估和开发航拍图像中的物体检测算法，具有尺度、方向和形状的多样性特点。