LACAS2K: A Large-Scale Aerial and Satellite Image Dataset for Image Super-Resolution of VHR Remote Sensing Imagery

LACAS2K is a large-scale dataset of aerial and satellite imagery designed for training and evaluating image super-resolution (SR) models, with a specific focus on very high-resolution (VHR) satellite imagery. The dataset is constructed with a standardized input spatial resolution of approximately 50 cm, aligning with the typical resolution of panchromatic (PAN) bands in modern VHR satellites.To support 2\u00d7 and 4\u00d7 SR tasks, the dataset provides high-resolution (HR) image patches of size 2048\u00d72048 (2K). The HR aerial images are at spatial resolutions of 25 cm and 12 cm, corresponding to scale factors of 2 and 4, respectively. Meanwhile, the HR satellite images are at 50 cm resolution for both scale factors. The corresponding low-resolution (LR) images are generated via bicubic downsampling to ensure consistency.Minor refinements were made to improve image quality in the released dataset. As a result, the results reported in our IEEE JSTARS publication may not be fully reproducible. However, the released version provides images of higher quality and is recommended for future research and benchmarking.Each subset contains 5,000 image pairs per scale factor, divided into training (0001\u20134000), validation (4001\u20134500), and test (4501\u20135000) sets in an 8:1:1 ratio. All images are provided in TIFF format to preserve data fidelity and avoid compression artifacts.