Science Wednesday: roll the Dice for change detection

To monitor the ways that people are managing agricultural lands, we are going to need to be able to detect changes in land cover.

In this new paper, three Digiscape alumni propose a new deep learning framework for semantic change detection from high-resolution remote-sensing imagery. The key insight: facilitating the use of relative attention between two convolution layers in order to fuse them. Validation with two building change detection datasets shows excellent performance: