Processed Hyperspectral Images and Ground-Truth Labels for Cross-Scene Hyperspectral Image Classification via Semantic-Guided Multi-Level Domain Alignment

This dataset contains the processed hyperspectral images and corresponding ground-truth label maps used in the three cross-scene classification tasks. In all ground-truth label maps, 0 denotes unlabeled pixels, while labels 1–4 represent the four shared land-cover classes: 1—roads, 2—grass, 3—trees, and 4—roofs.