Brussels / 3 & 4 February 2018


How DeepLearning can help to improve geospatial DataQuality , an OSM use case.

How DeepLearning, and semantic segmentation, can be an efficient way to detect and spot inconsistency in an existing dataset ? OpenStreetMap dataset took as an use case.

DataQuality is a must, but a gageure. And any technique on any help to improve DataQuality is then more than welcome.

Machine and DeepLearning can succeed to tackle some old issues, in a far more convenient and efficient way than ever before... For instance DeepLearning, with aerial imagery semantic segmentation can detect features from an aerial image and allow us to check dataset consistency.

In this presentation we will focus on how an OpenStreetMap subset dataset (for instance roads and buildings on an area), can be evaluated to produce a quality metric, and to spot areas where obvisously dataset is still far to be complete.

On a DataScience point of view, we wanna focus on:

  • DeepLearning vision, and specific Satellite imagery considerations (high and lower resolutions, multispectral dimensions, dataset aggregation...)
  • How to qualify a good enough labelled DataSet (to allow supervised learning)
  • PostgreSQL/PostGIS integration with Python ML/DL framework
  • Concrete solution for efficient treatments for wide coverages

On a OpenData, point of view we will consider, how this kind of solution could be integrated with OSM project Assurance Quality policy.


Photo of Olivier Courtin Olivier Courtin