航空图像中目标检测的大规模数据集DOTA

DOTA 数据集（全称 A Large-scale Dataset for Object DeTection in Aerial Images）是用于航拍图像中的目标检测的大型图像数据集。 它可用于发现和评估航拍图像中的物体。DOTA-v1.0 包含来自不同传感器和平台的 2806 幅航拍图像。 每个图像的尺寸在约 800×800 到 4000×4000 像素的范围内，并且包含呈现各种尺度、方向和形状的物体。 然后，这些 DOTA 图像由航空图像解释专家使用 15 个常见目标类别进行注释。 完全注释的 DOTA 图像包含 188,282 个实例，每个实例由任意（8 d.o.f.）四边形标记。 DOTA 数据集于 2017 年 11 月 28 日由武汉大学发布在 arXiv 上，之后于 2018 年 6 月发布在 IEEE 计算机视觉和模式识别会议（CVPR）上。DOTA-v1.0 数据集中的图像是从 Google Earth 收集的，其中一些是由卫星 JL-1 拍摄的，其他的是由中国资源卫星数据和应用中心的卫星 GF-2 拍摄的。

The DOTA dataset (full name: *A Large-scale Dataset for Object Detection in Aerial Images*) is a large-scale image dataset dedicated to object detection in aerial imagery, which is used to discover and evaluate objects within aerial images. DOTA-v1.0 contains 2806 aerial images from different sensors and platforms. The resolution of each image ranges from approximately 800×800 to 4000×4000 pixels, and includes objects with various scales, orientations and shapes. Subsequently, these DOTA images were annotated by aerial image interpretation experts using 15 common object categories. The fully annotated DOTA images contain 188,282 object instances, each marked by an arbitrary (8 d.o.f.) quadrilateral. The DOTA dataset was released on arXiv by Wuhan University on November 28, 2017, and later published at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) in June 2018. The images in the DOTA-v1.0 dataset were collected from Google Earth, some of which were captured by the JL-1 satellite, while others were captured by the GF-2 satellite from the China Resources Satellite Data and Application Center.

搜集汇总

背景与挑战

背景概述

DOTA数据集是一个专用于航空图像目标检测的大规模数据集，包含2806幅高分辨率航拍图像，覆盖15个常见目标类别。数据集具有约188,282个实例的详细注释，每个实例以任意四边形标记，适用于多尺度、多方向的目标检测任务，于2017-2018年发布，常用于学术研究和算法评估。