NeurIPS: Tackling Climate Change with Machine Learning

This year I’ve published a study in close collaboration with great students at DTU. We present a benchmark dataset for the detection and localization of residential PV systems from aerial imagery.

Among the surprising findings is the poor generalization of solar recognition classifiers across geographical regions that semantically looks very similar.

Our paper and our presentation can be found here. A big thanks to the organizers at ClimateChange.ai for an interesting workshop.