SpaceNet MVOI
收藏资源简介:
SpaceNet MVOI是一个开放源代码的多视角高空影像数据集,由In-Q-Tel CosmiQ Works创建。该数据集包含60,000张高空影像,覆盖美国乔治亚州亚特兰大及其周边地区的665平方公里地理范围,并标注了126,747个建筑物轮廓。数据集的影像来自Maxar WorldView-2卫星,具有27个不同的视角,覆盖了从-32.5°到54.0°的广泛视角范围。该数据集旨在评估视角变化对模型性能的影响,特别是在动态场景如自然灾害中的应用。SpaceNet MVOI适用于建筑物检测、模型泛化能力测试以及分辨率变化对模型性能的影响分析,为计算机视觉领域提供了重要的基准数据。
SpaceNet MVOI is an open-source multi-view overhead imagery dataset created by In-Q-Tel CosmiQ Works. This dataset contains 60,000 overhead images covering a 665-square-kilometer geographic area over Atlanta, Georgia, USA and its surrounding regions, with 126,747 annotated building footprints. The imagery in the dataset is sourced from the Maxar WorldView-2 satellite, featuring 27 distinct viewpoints spanning a wide angular range from -32.5° to 54.0°. This dataset is designed to evaluate the impact of viewpoint variation on model performance, particularly for applications in dynamic scenarios such as natural disasters. SpaceNet MVOI is suitable for building detection, model generalization capability testing, and analysis of the impact of resolution changes on model performance, providing important benchmark data for the computer vision field.



