UAV RGB Imagery, Tree-Crown Masks, and Tree-Height Maps for Haloxylon ammodendron Crown Segmentation and Height Estimation

This dataset contains UAV RGB images, tree-crown segmentation masks, and processed tree-height raster labels for individual crown segmentation and tree-height estimation of Haloxylon ammodendron. The uploaded files include two compressed archives. The file VOCdevkit.zip contains the tree-crown segmentation dataset organized in VOC format, including UAV RGB image patches, segmentation masks, and training/validation split files. The file TreeHeightDataset.zip contains the tree-height regression dataset, including RGB image patches, crown masks, processed height maps, and training/validation split files. This dataset was used to develop and evaluate the LSNet-BiFoST framework, which integrates a tree-crown segmentation branch and a mask-guided tree-height regression branch. The tree-crown segmentation masks are used not only for semantic segmentation evaluation but also as spatial guidance for tree-height regression, thereby reducing the influence of background interference on tree-height estimation. The dataset is intended for research in UAV remote sensing, individual tree-crown segmentation, tree-height estimation, deep learning, semantic segmentation, and ecological monitoring in arid environments.