2026-09-01 –, Conference Management Room1
Introducing a deep learning-based rooftop solar model using open remote sensing data and OpenStreetMap. We will explain the design of the model and demonstrate its use in the Climate Action Navigator.
How dependent is your town on the electricity grid? How many buildings are supplied by rooftop solar panels? How much unused potential is there to leverage the power of the sun?
We built a deep learning model using FOSS4G and open remote sensing data for Germany. With this model, you can detect which buildings have rooftop solar panels at a neighbourhood level. The input data is orthophotos and OpenStreetMap building footprints, which we feed into a 4-channel image classification model.
The results of the model are visualised in the Rooftop Solar assessment tool of the Climate Action Navigator (https://climate-action.heigit.org) from HeiGIT (https://heigit.org).
In this talk, we will demonstrate our results through our assessment tool. We will also explain the design of our model and how we used OpenStreetMap tagging to significantly speed up the creation of training data for our supervised learning approach.
We used open source software including Python, pytorch, and supporting libraries for developing and training our model. We also used free and open data from OpenStreetMap and digital orthophotos for Germany.
Our project, the Climate Action Navigator, is also FOSS4G (GNU v3): https://climate-action.heigit.org
I make my conference contribution available under the CC BY 4.0 license. The conference contribution comprises the abstract, the text contribution for the conference proceedings, the presentation materials as well as the video recording and live transmission of the presentation:
Navigating and understanding the growing world of data and technologies is becoming increasingly complex. I strive at connecting data with strategy, helping to understand and explain the stories we can uncover behind data to support good decision making and lead to a more sustainable future.
