HeiGIT contributes to this year´s FOSSGIS conference with two sessions:
Talk: Enrichment of Road Data Using Deep Learning Methods and Mapillary Images (Anreicherung von Straßendaten mittels Deep-Learning-Methoden und Mapillary Bildern)
Friday, 27/03/2025, 9:00-9:20 am, HS3 (S1) – Benjamin Herfort, Sukanya Randhawa
In their 20-minute talk session, Benjamin Herfort and Sukanya Randhawa explain how they created a global dataset of road properties (‘paved’ or ‘unpaved’) using AI methods based on 105 million images from Mapillary. This approach combines SWIN-transformer-predictions of the road surface with CLIP filtering of low-quality images. The dataset covers 36% of the world’s road network, with higher coverage in cities, while highlighting global disparities. This dataset can be used in urban planning, disaster management and logistics. Its potential uses will be described and explained in more detail in this talk.
Workshop: OSM-Transform – Powerful OSM Data Preparation for Routing Engines (OSM-Transform – Leistungsstarke OSM-Datenvorbereitung für Routing Engines)
Friday, 28/03/2025, 11:10-12:40 am, WS2 (107) – Julian Psotta, Jakob Schnell, Till Frankenbach
Julian Psotta, Jakob Schnell and Till Frankenbach will introduce OSM-Transform, the preprocessor of the open source routing engine openrouteserviece (ORS) in their 90-minute workshop session. Participants will learn how to process OSM data for efficient routing and how to integrate elevation data and country boundaries from various sources and resolutions. Using simple examples, the hosts will show how a routing engine can be set up in just a few steps using the processed data, thus enabling future analyses. The functions covered in the workshop are data filtering, enrichment, elevation data, specialized algorithms, easy setup and further processing. The future functions global-high resolution elevation data, timezones information and generic data enrichment will also be discussed.
We are looking forward to welcoming and discussing with you at these interesting sessions!
