GeoPython 2023

The Humanitarian OpenStreetMap Team (HOT) utilizes OpenStreetMap (OSM) for its humanitarian actions and community development. The spatial location of buildings is essential information in terms of disaster resilience. During the OMDENA challenge, 50 machine learning engineers developed an AI-assisted mapping tool supporting the HOT mappers in delineating building footprints based on aerial images from OpenAerialMap (OAM).

Based on the input data provided by HOT, the project team developed a workflow for preprocessing aerial images as well as the building footprints. Several architectures of Convolutional Neural Networks (CNN) were tested during the challenge. As the provided input data was limited, the final approach was to apply the pre-trained model from the RAMP project. This deep neural network was trained with large amounts of satellite images and was used for transfer learning.

One major outcome of the challenge was the development of the HOTLib library as a reference implementation of the planned workflow. This Python library supports HOT in the preprocessing of the input data, the inference step based on the RAMP model, and the post-processing of the prediction results.

Additional information:
- https://omdena.com/projects/mapping-tool-for-disaster-management
- https://omdena.com/blog/ai-assisted-collaborative-mapping-tool-with-humanitarian-openstreetmap
- https://www.hotosm.org/tech-blog/hot-tech-talk-open-ai-challenge
- https://www.hotosm.org/tech-blog/hot-tech-talks-fair