Is deep learning the new agent for map generalization?

ABSTRACT

The automation of map generalization has been keeping researchers in cartography busy for years. Particularly great progress was made in the late 90s with the use of the multi-agent paradigm. Although the current use of automatic processes in some national mapping agencies is a great achievement, there are still many unsolved issues and research seems to stagnate in the recent years. With the success of deep learning in many fields of science, including geographic information science, this paper poses the controversial question of the title: is deep learning the new agent, i.e. the technique that will make generalization research bridge the gap to fully automated generalization processes? The paper neither responds a clear yes nor a clear no but discusses what issues could be tackled with deep learning and what the promising perspectives. Some preliminary experiments with building generalization or data enrichments are presented to support the discussion.

RÉSUMÉ

Les chercheurs en cartographie essaient d’automatiser le processus de généralisation cartographique. Des progrès importants ont été faits à la fin des années 90 avec l’utilisation du paradigme multi-agents introduit par Anne Ruas. Mais même si des processus automatiques de généralisation sont aujourd’hui utilisés dans plusieurs agences nationales de cartographie, il reste encore de nombreuses situations où on ne sait pas bien généraliser automatiquement, et la recherche dans le domaine semble stagner ces dernières années. Face au succès grandissant des techniques d’apprentissage profond dans de nombreux domaines, y compris liés à la cartographie, cet article pose la question controversée du titre : est-ce que l’apprentissage profond va nous permettre de faire un bond en avant vers l’automatisation complète à la manière du paradigme multi-agent ? L’article ne répond pas clairement à la question, mais discute du potentiel des techniques d’apprentissage profond en généralisation cartographique. Des expérimentations préliminaires sont présentées pour illustrer cette discussion, sur la généralisation des bâtiments et la détection des échangeurs routiers.

KEYWORDS:

• Map generalization
• machine learning
• deep learning

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Guillaume Touya is vice-head of the LASTIG lab, IGN France (the French mapping agency), and a senior researcher in the GEOVIS team of LASTIG. He holds a PhD and habilitation in GI science from Paris-Est University. His research interests focus on automated cartography, map generalization and volunteered geographic information. He currently leads the MapMuxing (https://mapmuxing.ign.fr) research project on mixing cartography and human–computer interaction issues. He is the chair of the ICA commission on map generalization and multiple representation.

Xiang Zhang is an associate professor at the School of Resource and Environmental Science, Wuhan University, China. He received his PhD in Cartography and Geographic Information Science from the University of Twente, the Netherlands in 2012. His research interest includes integration, analysis, generalization, and visualization of spatiotemporal data. In his previous work, methods from computational geometry, pattern recognition, and machine learning were extended to tackle fundamental and applied problems in GIScience.

Imran Lokhat has been a developer at the LASTIG lab, IGN France (the French mapping agency) since 2014. He works on the open source platforms developed in the lab: GeOxygene (https://github.com/IGNF/geoxygene), iTowns (https://github.com/itownsResearch), and CartAGen (https://github.com/IGNF/CartAGen). He recently gained interest in deep learning techniques for GI science.

ORCID

Guillaume Touya http://orcid.org/0000-0001-6113-6903

Notes

1. https://github.com/IGNF/CartAGen

Additional information

Funding

Xiang Zhang was supported by the National Natural Science Foundation of China (grant 41671384) and the National Key Research and Development Program of China (grant 2017YFB0503500).