MAI (Multi-scene Aerial Image)

MAI 是用于在单个航拍图像中进行多场景识别的数据集。它由来自 Google 地球图像的 3,923 张标记的大型图像组成，覆盖美国、德国和法国。每张图像的尺寸为 512 × 512，空间分辨率从 0.3 m/pixel 到 0.6 m/pixel 不等。在捕获航拍图像后，从总共 24 个场景类别中为每个图像手动分配多个场景级标签，包括围裙、棒球、海滩、商业、农田、林地、停车场、港口、住宅、河流、储罐、海洋、桥, 湖, 公园, 迂回, 足球场, 体育场, 火车站, 作品, 高尔夫球场, 跑道, 稀疏灌木和网球场

MAI is a dataset for multi-scene recognition in single aerial images. It comprises 3,923 manually labeled large-scale images sourced from Google Earth, covering the United States, Germany and France. Each image has a dimension of 512 × 512 pixels, with spatial resolutions ranging from 0.3 m/pixel to 0.6 m/pixel. After capturing these aerial images, multiple scene-level labels were manually assigned to each image from a total of 24 predefined scene categories, including apron, baseball field, beach, commercial area, farmland, woodland, parking lot, port, residential area, river, storage tank, ocean, bridge, lake, park, roundabout, football field, stadium, railway station, works, golf course, runway, sparse shrubs and tennis court

搜集汇总

背景与挑战

背景概述

MAI是一个用于航拍图像多场景识别的数据集，包含3,923张512×512像素的图像，空间分辨率在0.3到0.6米/像素之间，覆盖美国、德国和法国，每张图像手动标注了24个场景类别中的多个标签。该数据集由学术机构于2021年发布，适用于计算机视觉中的场景识别研究，支持多标签分类任务。