We observe that satellite imagery is a powerful source of information as it contains more structured and uniform data, compared to traditional images. Although computer vision community has been accomplishing hard tasks on everyday image datasets using deep learning, satellite images are only recently gaining attention for maps and population analysis. This workshop aims at bringing together a diverse set of researchers to advance the state-of-the-art in satellite image analysis.

To direct more attention to such approaches, we propose DeepGlobe Satellite Image Understanding Challenge, structured around three different satellite image understanding tasks: road extraction, building detection, and land cover classification.

Participants can submit to a single track or multiple tracks. We encourage all submissions including novel techniques, approaches under review, and already published methods. The first three places of each track will receive prizes. The winner in each track will give an oral presentation in DeepGlobe CVPR 2018 workshop, and the top five submissions in each track will be invited for a poster presentation. We will also publish selected papers in the proceedings of the workshop.