Machine learning against natural disasters: segmentation of satellite images using deep learning

| |

The lecture reveals what models, data points – and logic should be used in machine learning data analysis

One of the keys to the prevention and management of natural disasters is an up – to – date knowledge of the built environment. This is a particular problem in developing countries, which is why DrivenData Labs and the Global Facility for Disaster Reduction and Recovery (GFDRR) have launched an open competition to solve the problem. The Starschema team took part in this competition, and the lessons of the competition were presented on Tuesday, March 31, 2020, CountingNews free! In a presentation at a meeting, two of the company's data scientists, Balázs Zempléni and Benedek Tóth, visited

What are the usual methods for processing satellite images and recognizing buildings? What model is worth teaching, on what data and exactly what logic? How can the teaching data and the finished prediction be manipulated to achieve better results? Where are the technological limitations of the process and what kind of project structure does it require to overcome them? From the beginning to the end of the lecture, he reports on an applied deep learning solution, both on the theoretical background and on implementation issues.

Machine learning against natural disasters – Balázs Zempléni, Benedek Tóth (Starschema)

More videos

Previous

A crucial year awaits Huaweire, who has switched to survival mode

The bite of DIGI Mobil hurts the market less than expected

Next