Satellite Images Analysis

Hundreds of thousands of hectares of natural areas are destroyed in Ukraine every year by wildfires. This natural disaster causes huge and sometimes irreparable damage to nature. Billions of insects, animals, birds as well as seeds and roots of plants, located in the upper layers of the soil die in the fire. It takes a lot of time for natural ecosystems to recover and some of their components can not be recovered without human assistance. Manually it takes an expert from 8 to 20 to analyze one satellite image. So we aim to create an AI application that can detect burned areas on satellite images taken by various satellites with different sensors onboard.

Goals:  

* investigate the root cause of wildfires 

* automate the process of wildfires detection using machine learning algorithms

Instructor's Bio

Ruslan Partsey, Computer Vision researcher in ML lab at UCU


Graduated from Ivan Franko National University with a Bachelor’s Degree in Applied Mathematics. Used to work as a python software engineer for 1.5 years. Joined UCU ML Lab 2 months ago. Currently doing the Master’s Degree in Data Science at Ukrainian Catholic University.

Face-reenactment or Neural Avatar Generation

Artificial avatar is a brand-new notion of the last few years with many research papers published every month. This technology has a myriad of applications in the industries of cinematography, mass-media, AR/VR, gaming, to name just a few. It can accelerate the process of dubbing in multiple languages, enhance teleconferencing experience with the 3D avatar of a person, personalize commercials even more. Neural networks that can produce such images and videos, in general, are heavy. Retraining them every time for another person is very costly. So we introduce the method, that is agnostic to the choice of avatar.

We acknowledge that such technology could be used with offensive purposes and exploits ethical questions. Nowadays, an average human cannot tell the difference between real or fake face. With this technology, counterfeit videos of politicians, signers, and other public figures could be produced. Therefore, this technology should be used in limited conditions. Access to it should be granted for the purposes, that do not violate human rights. We hope that our work will help researchers improve models that can detect such synthetic content.

IT giants like Facebook, Amazon, Microsoft are aware of the possible misuse of this technology. Recently they have conducted a Deepfake detection challenge https://deepfakedetectionchallenge.ai/ with a prize of $1 million for the best classification model.

Instructors' Bio

Marian Petruk, R&D Engineer at SoftServe

Currently, Marian is finishing a Computer Science program (BSc) at the Ukrainian Catholic University. He is working as an R&D Engineer at SoftServe and also as a teaching assistant in the Programming Fundamentals course. Marian graduated from Lviv Academic Gymnasium and Computer Academy "STEP".


Ivan Kosarevych, R&D Engineer at SoftServe  

Currently, Ivan is working as an R&D Engineer at SoftServe. Meanwhile, he is a Senior year Computer Science (BSc) student at Ukrainian Catholic University. 

Webinar

  • 01

    Satellite Images Analysis & Face-reenactment or Neural Avatar Generation

    • Webinar Link