Today, our employee, Hanna Yailymova, at the annual forum of the international Earth observation group GeoWeek 2021 presented the project «Methods of deep learning for the classification of land cover and land use» under the GEO-GEE program.
The project was nominated by Google Earth Engine, and we received licensed access to conduct research on the Google Cloud Platform. Our main goal is to adapt and improve existing in-depth learning methods for mapping the earth using satellite data (up to 10 m) at the national level and to improve the digitalization of the economy and decision-making processes in Ukraine.
As a result, we obtained a classification map according to our deep learning algorithm based on the time series of satellite data from the GEE cloud platform. The overall accuracy of the received maps is 95%.
We do not stop there – there are still many areas for expansion and improvement that we need to cover.