change detection dataset

It primarily covers reconstruction areas in the years following an earthquake, focusing on changes in buildings. In the experiments, the original images were cropped into 256 × 256 pixel patches, and the dataset was split into training, validation, and test sets in an 8:1:1 ratio.  LEVIR-CD dataset: Sourced from Google Earth, with a spatial resolution of 0.5 meters and an original image size of 1024×1024 pixels.

This dataset exhibits significant land-use changes, particularly including various types of building changes.In the experiment, non-overlapping cropping was used to generate 256×256 image patches, resulting in a total of 7,120 training patches, 1,024 validation patches, and 2,048 test patches.  CLCD Dataset: Composed of 2017 and 2019 Gaofen-2 satellite images, it contains 600 pairs of 512×512 dual-temporal samples with a spatial resolution of 0.5–2 meters.The changes in land cover are primarily related to cropland, while also covering various land features such as buildings, roads, lakes, and bare ground.