Learning to Interpret Satellite Images Using Wikipedia
Abstract
Despite recent progress in computer vision, fine-grained interpretation of satellite images remains challenging because of a lack of labeled training data. To overcome this limitation, we construct a novel dataset called WikiSatNet by pairing geo-referenced Wikipedia articles with satellite imagery of their corresponding locations. We then propose two strategies to learn representations of satellite images by predicting properties of the corresponding articles from the images. Leveraging this new multi-modal dataset, we can drastically reduce the quantity of human-annotated labels and time required for downstream tasks. On the recently released fMoW dataset, our pre-training strategies can boost the performance of a model pre-trained on ImageNet by up to 4.5% in F1 score.
Cite
Text
Uzkent et al. "Learning to Interpret Satellite Images Using Wikipedia." International Joint Conference on Artificial Intelligence, 2019. doi:10.24963/IJCAI.2019/502Markdown
[Uzkent et al. "Learning to Interpret Satellite Images Using Wikipedia." International Joint Conference on Artificial Intelligence, 2019.](https://mlanthology.org/ijcai/2019/uzkent2019ijcai-learning/) doi:10.24963/IJCAI.2019/502BibTeX
@inproceedings{uzkent2019ijcai-learning,
title = {{Learning to Interpret Satellite Images Using Wikipedia}},
author = {Uzkent, Burak and Sheehan, Evan and Meng, Chenlin and Tang, Zhongyi and Burke, Marshall and Lobell, David B. and Ermon, Stefano},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2019},
pages = {3620-3626},
doi = {10.24963/IJCAI.2019/502},
url = {https://mlanthology.org/ijcai/2019/uzkent2019ijcai-learning/}
}