The Syrian conflict has been the biggest humanitarian crisis in recent time taking the live of millions of people and destroying urban centers and infrastructures in the country. Due to restricted access, security issue as well as dysfunction of main institutions the scale of the destruction remains to be estimated accurately. This work aims at giving a clue on the later. Building on satellite imagery and the huge work of the UNITAR-UNOSAT (the UN satellite imagery program) we construct a classifier that recognizes destructed infrastructures and buildings in main Syrian urban centers. Based on this algorithm we then propose a mapping of destruction for several cities in Syria were information is still very scarce. Further improvments are finally proposed.
Instructions to reproduce our results :
We have numbered our notebooks so that the user can easily find their way around in the required order. The main notebooks are gathered in the folder notebooks, while the part concerning transfer learning on the EuroSAT dataset in set aside in the folder euroSAT.