Amhara Region WorldView-CNN LCLU training imagery and labels, version 1.0

This dataset provides paired WorldView-derived top-of-atmosphere imagery and land-cover/land-use training labels used for CNN-based LCLU mapping in Amhara Region, Ethiopia. The dataset supports training, transfer learning, and evaluation workflows for very-high-resolution remote-sensing classification in fragmented smallholder landscapes. The archive includes multispectral WorldView image rasters, aligned categorical label rasters, class definitions, band-order documentation, file manifests, image-label pair metadata, and checksums. Label rasters were aligned to the exact pixel grid of their corresponding input imagery using nearest-neighbor resampling. The label values are: 0 = Crop, 1 = Tree/Shrub, 2 = Grass, 3 = Built-up, 4 = Water, and 15 = NoData / ignore / unlabeled. Commercial WorldView and GeoEye source imagery is subject to access and redistribution restrictions. This record includes only redistribution-permitted derived imagery and label products. Users should cite this Zenodo record when using the training imagery or labels.