Health + Life Science Alliance Heidelberg Mannheim

This project has successfully developed two GeoAI models designed to extract detailed information about urban greenspaces from street view imagery. The first model employs semantic segmentation techniques to identify pixels associated with trees, low-lying vegetation, and grass. The second model uses image regression to estimate vegetation diversity and naturalness. Both models have been validated and meet established accuracy benchmarks. We analyzed approximately 250,000 Google Street View images from the Mannheim region, where data on mental health is available, using these models. Additionally, we devised a method to incorporate images contributed by volunteers from Mapillary. In the project's concluding stage, we are developing statistical models to investigate the relationships between the structure, diversity, and naturalness of urban greenspaces and mental health.