loading page

DeepLandforms: A Deep Learning Computer Vision toolset applied to a prime use case for mapping planetary skylights
  • +1
  • Giacomo Nodjoumi,
  • Riccardo Pozzobon,
  • Franceso Sauro,
  • Angelo Pio Rossi
Giacomo Nodjoumi
Jacobs University Bremen

Corresponding Author:[email protected]

Author Profile
Riccardo Pozzobon
Università degli Studi di Padova
Author Profile
Franceso Sauro
University of Bologna
Author Profile
Angelo Pio Rossi
Jacobs University Bremen
Author Profile


Thematic map creation is a process that implies several steps to be accomplished regardless of the type of the map to be produced, starting from data collection, through data exploitation and ending with maps publication as print, image, and GIS format. An example are geological, lithological, and geomorphological maps, in which most of the highest time-consuming tasks are those related to the discretization of single objects by identifying a set of unique characteristics that describe uniquely those objects. Commonly these tasks introduce interpretative biases due to the different experience of the mappers who's performing it. In this setting, Deep Learning Computer Vision techniques could play a key-role but lack the availability of a complete set of tools specific for planetary mapping. The aim of this work is to develop a comprehensive set of ready-to-use tools for landforms mapping, in which users have full access to the workflow and over all the processes involved, granting complete control and customization capabilities. In this work are presented both the developed tools and the approach that has been used and that is based on consolidated Deep Learning methodologies and open-source libraries commonly applied in other fields of Computer Vision. The toolset and the approach presented have been tested in the science case of mapping sinkhole-like landforms on Mars and results are presented.