Style transfer in Map Design
Recent advances in deep learning have facilitated the exchange of styles and textures between input images to create unique synthesised outputs. In this project I assessed the applicability of neural style transfer to cartography and evaluated to what degree emotions attached to input images can be preserved in maps co-created by human and algorithm. As a source of emotions I utilized personal paintings created during a workshop with international artists. The neural style transfer was used as a tool to transfer the characteristics of the artworks onto the map.
The content map input was created in Mapbox Studio and consisted of roads, buildings and land cover classes represented in light colour scheme. The handmade style inputs were collected at the School of Machines, Making & Make-Believe in Berlin during the Autonomous Generative Spirit Programme in August 2018. The participants were asked to reflect on their emotions about Berlin and to visualize them in the painting process, focussing on conveying the unique style of their feelings by the choice of colours, lines, and painted structures. Based on their feelings, 16 textures in both abstract and realistic styles were created and scanned. The last input image combined all personal paintings together to express a collaborative, emotional image of the city.
The personal emotional images were projected to the base map using the neural style transfer. The created abstract maps of the city are vivid and valid for single person or an artist’s group in a particular lifetime and location. You can read more about this approach in our paper.