PanScale is a benchmark dataset for large-scale inference and capability assessment, released in 2026 by the Chinese Academy of Sciences in conjunction with the University of Science and Technology of China and the Hong Kong University of Science and Technology. Related research papers include... Cross-Scale Pansharpening via ScaleFormer and the PanScale BenchmarkThe aim is to improve the model's ability to fuse and reconstruct under cross-resolution conditions. This dataset contains 7,559 pairs of multispectral (MS) and panchromatic (PAN) images in 8-bit TIFF format. It covers multiple subsets of the Jilin, Landsat, and Skysat datasets, and extends to cross-scale versions such as fjilin, flandsat, and fskysat, supporting systematic evaluation of scenes from the same scale to multiple scales (up to 4.0x). Each dataset consists of a 4-channel multispectral image and a 1-channel panchromatic image, and is widely used for panchromatic sharpening model training, cross-scale generalization analysis, and remote sensing image processing research.

Data fields:

• ms: Multispectral image, 4-channel TIFF format
• pan: Pancolor image, 1-channel TIFF format
• id: An integer type representing a unique identifier for an image pair (file terminology stem).