On December 14, AeronetLab of CDISE (Skoltech) took part in the GIS Tech Russia conference, organized by the University of Innopolis. Georgy Potapov presented on the topic “How to Quantify People from Space.”
It showed how to collect building and population data in a feasible and time-efficient way using large quantities of satellite imagery and processing algorithms powered by neural networks. Such applications are useful for location intelligence and business planners, since the companies in these markets usually have limited time or financial resources to carry out detailed field reports.
The method is based on the current research in progress within “Geoalert,” which is the platform for automatic analysis of Earth observation data provided by the world’s leading satellite operators.
Kirov dataset: 20K+ buildings extracted and classified by Geoalert model using the imagery from Digitalglobe
At the Geoanalytics workshop, GIS Tech Russia, Innopolis
Additionally, the “GeoSolvers” team with AeronetLab research engineer, Pavel Yakubovsky, took second place in the ongoing hackathon. The task was to recognize buildings in aerial imagery with a 10cm resolution. The team that took first place was “AddHueToSatur.”
Congratulations and thanks to all participants! Organizing such competitions allows us to better understand and formulate the requirements for open datasets and benchmarks provided by the consortium involving Skoltech, with the goal of conducting research in the field of machine analysis of remote sensing data.
Paricipants of GIS Tech Hack
+7 (495) 280 14 81