ChangeAnywhere-100K

该数据集是迄今为止最大的合成变化检测数据集，包含了使用ChangeAnywhere方法生成的10万对变化检测样本。该方法利用了语义潜在扩散模型，能够产生具有现实变化场景及非变化区域多样性的样本。这些样本不仅展示了ChangeAnywhere方法在生成具有丰富多样性和语义标注的双时相数据集方面的能力，而且数据集的规模达到了10万对变化检测样本。该数据集主要针对的任务是遥感图像中的变化检测。

This dataset represents the largest synthetic change detection dataset to date, encompassing 100,000 pairs of change detection samples generated via the ChangeAnywhere method. This approach leverages semantic latent diffusion models, allowing for the creation of samples that exhibit realistic change scenarios and diverse unchanged regions. These samples not only showcase the ability of the ChangeAnywhere method to generate bi-temporal datasets with abundant diversity and semantic annotations, but also highlight the substantial scale of this dataset, which totals 100,000 pairs of change detection samples. This dataset is primarily tailored for the task of change detection in remote sensing imagery.