This repository contains the implementation for our publication "Map-Repair: Deep Cadastre Maps Alignment and Temporal Inconsistencies Fix in Satellite Images", IGARSS 2020. If you use this implementation please cite the following publication:

@inproceedings{zorzi2020map,
  title={Map-repair: Deep cadastre maps alignment and temporal inconsistencies fix in satellite images},
  author={Zorzi, Stefano and Bittner, Ksenia and Fraundorfer, Friedrich},
  booktitle={IGARSS 2020-2020 IEEE International Geoscience and Remote Sensing Symposium},
  pages={1829--1832},
  year={2020},
  organization={IEEE}
}

Explanatory video of the approach:

• cuda 10.2
• pytorch >= 1.3
• kornia
• opencv
• gdal

After installing all of the required dependencies above you can download the pretrained weights from here.

Unzip the archive and place the content in the main maprepair folder. The folder saved_models contains the pretrained weights both for MapRepair and the regularization network.

Modify variables.py accordingly, then run the prediction issuing the command

python predict.py

Modify variables.py accordingly, then run the training issuing the command

python train_net.py