Company-Project:
IIASA/Brockmann consult - Picture Pile/Crowd2Train
Description:
Using the Crowd toTrain platform to generate crop type maps based on crowdsourced classification of Streetlevel Photography

A workflow with an end to end use case will be presented which allows to use the EuroData Cube Jupiter Notebook to generate crop type maps from crowdsourced street level photography.
Based on a specified region of interest which contains street level photography (Google Streetview or Mapillary images), segments based on Sentinel 2 images corresponding to the same growing season as the Streetlevel photographs available, will be generated. The centroid of each segment along roads will provide the input data for the Streetview photograph extraction which will go into the picture pile swiping tool which collects crop type information. This process allows the labelling of the segments which can then be used in a machine learning crop type classification algorithm. The final result of the workflow will be a crop type map of the specified region.