GeoText/GeoText-1652
收藏资源简介:
GeoText-1652数据集是一个用于地理位置识别和自然语言引导无人机任务的基准数据集。数据集包含来自无人机、卫星和地面摄像头的图像,分为训练集和测试集。训练集和测试集均包含图像、全局描述、边界框文本对和建筑物编号。数据集的目录结构包括训练和测试图像文件夹,以及包含注释和元数据的JSON文件。训练集和测试集的图像分别来自33所和39所不同的大学,且两者之间没有重叠。数据集的设计旨在支持地理定位识别模型的开发和测试,提供来自多个视角的图像。
The GeoText-1652 dataset is designed for natural language-guided drones and spatial relation matching tasks. The dataset includes training and test sets with images, global descriptions, bbox-text pairs, and building numbers. The training and test sets are divided into three platforms: drone, satellite, and ground cameras. The dataset also includes JSON files with annotations for both training and testing, which are essential for tasks like image classification, object detection, and scene parsing. The README also describes the structure of the dataset, including directories and files, and provides an example entry from the `train.json` file to illustrate the format of the annotations.




