LEVIR-CD
收藏资源简介:
LEVIR-CD 是一个新的大规模遥感建筑变化检测数据集。引入的数据集将成为评估变化检测 (CD) 算法的新基准,尤其是基于深度学习的算法。 LEVIR-CD 由 637 个非常高分辨率(VHR,0.5m/像素)Google Earth (GE) 图像块对组成,大小为 1024 × 1024 像素。这些时间跨度为 5 到 14 年的双时相图像具有显着的土地利用变化,尤其是建筑增长。 LEVIR-CD涵盖别墅住宅、高层公寓、小型车库和大型仓库等各类建筑。在这里,我们关注与建筑相关的变化,包括建筑增长(从土壤/草地/硬化地面或在建建筑到新建筑区域的变化)和建筑衰退。这些双时相图像由遥感图像解释专家使用二进制标签(1 表示变化,0 表示不变)进行注释。我们数据集中的每个样本都由一个注释器进行注释,然后由另一个注释器进行双重检查以产生高质量的注释。完整注释的 LEVIR-CD 总共包含 31,333 个单独的变更构建实例。
LEVIR-CD is a novel large-scale remote sensing building change detection dataset. This introduced dataset will serve as a new benchmark for evaluating change detection (CD) algorithms, especially deep learning-based ones. LEVIR-CD consists of 637 pairs of very high-resolution (VHR, 0.5 m/pixel) Google Earth (GE) image patches with a size of 1024 × 1024 pixels. These bi-temporal images, spanning a time range of 5 to 14 years, exhibit significant land use changes, particularly building growth. LEVIR-CD covers various types of buildings such as villas, high-rise apartments, small garages, and large warehouses. Herein, we focus on building-related changes, including building growth (changes from soil/grassland/hardened ground or under-construction buildings to new building areas) and building decline. These bi-temporal images were annotated by remote sensing image interpretation experts using binary labels, where 1 indicates a change and 0 indicates no change. Each sample in our dataset was first annotated by one annotator, then double-checked by another annotator to generate high-quality annotations. The fully annotated LEVIR-CD contains a total of 31,333 individual changed building instances.

- LEVIR-CD数据集首次发表,由Hao Chen等人提出,旨在为遥感图像变化检测提供一个标准化的数据集。
- LEVIR-CD数据集在多个遥感图像变化检测竞赛中被广泛应用,成为评估算法性能的重要基准。
- LEVIR-CD数据集的扩展版本发布,增加了更多的图像对和变化类别,进一步提升了数据集的多样性和实用性。
- 1LEVIR-CD: A Remote Sensing Building Change Detection DatasetWuhan University · 2020年
- 2A Deep Learning-Based Change Detection Method for Remote Sensing Images Using the LEVIR-CD DatasetWuhan University · 2021年
- 3Change Detection in Remote Sensing Images Using a Dual-Branch Network with the LEVIR-CD DatasetNanjing University of Information Science & Technology · 2022年
- 4A Comparative Study of Change Detection Methods on the LEVIR-CD DatasetUniversity of Electronic Science and Technology of China · 2022年
- 5Enhancing Change Detection in Remote Sensing Images with Attention Mechanisms Using the LEVIR-CD DatasetBeijing Normal University · 2023年



