mundialis has developed a flexible AI framework for differentiating surface classes from freely available aerial images, which we will present in two webinars. This first part highlights a practical example from wastewater management: we explain our approach, from the preparation of training data and the selection of a suitable AI architecture to the development of post-processing routines. The AI model we have developed is able to reliably and accurately distinguish between paved areas relevant to wastewater management, such as ‘public traffic areas’, ‘roof areas’, ‘private clean areas’ and ‘private dirty areas’, thus providing a comprehensive basis for the quantitative and qualitative assessment of precipitation runoff.
View the webinar recording.