AID multilabel dataset is produced from AID scene dataset. Details are introduced in this paper. In total, there are 3000 aerial images (2400 training and 600 test) and 17 object labels. The size of each image is 600 x 600 pixels. The proposed dataset is now available on Google Drive. Notably, in each set, images are sorted by their scene categories, defined in AID scene dataset.
If you find this dataset is useful, please kindly cite the following:
@article{hua2019relation,
title={Relation Network for Multi-label Aerial Image Classification},
author={Hua, Yuansheng and Mou, Lichao and Zhu, Xiao Xiang},
journal={IEEE Transactions on Geoscience and Remote Sensing},
year={DOI:10.1109/TGRS.2019.2963364}
}