ChangeNet
收藏资源简介:
ChangeNet数据集是由中国科学技术大学和阿里巴巴集团联合创建的,专注于多时相非对称变化检测的大型数据集。该数据集包含31,000对多时相图像,覆盖100个中国城市的复杂场景,并具有6个像素级标注类别。ChangeNet不仅适用于二值变化检测(BCD)和语义变化检测(SCD)任务,还特别强调了实际应用中的视角和形状畸变,旨在推动变化检测算法的实际应用。数据集的创建过程涉及从WayBack平台获取2014至2022年的0.3米分辨率图像,并通过高效的标注策略进行处理,以确保高质量的标注和较低的成本。ChangeNet的应用领域包括土地资源管理、城市化监测和土地退化评估等,旨在解决现有数据集在数量、时相和实用性方面的不足。
The ChangeNet dataset is a large-scale benchmark dataset jointly developed by the University of Science and Technology of China and Alibaba Group, specializing in multi-temporal asymmetric change detection. It comprises 31,000 pairs of multi-temporal images, covering complex scenarios across 100 Chinese cities, and includes 6 pixel-level annotated categories. Beyond supporting Binary Change Detection (BCD) and Semantic Change Detection (SCD) tasks, ChangeNet particularly highlights perspective and shape distortions encountered in real-world applications, with the objective of advancing the practical deployment of change detection algorithms. The dataset construction process involves acquiring 0.3-meter resolution imagery from the WayBack platform spanning 2014 to 2022, and leveraging efficient annotation workflows to ensure high-quality annotations while minimizing annotation costs. Application scenarios of ChangeNet cover land resource management, urbanization monitoring, land degradation assessment and other fields, aiming to address the shortcomings of existing datasets in terms of scale, temporal coverage and practical usability.




