[Humans of FDL] Marc Russwurm
Meet EO Researcher Marc Russwurm
We pulled Marc away from the FDL Europe lab in Oxford just long enough for a quick conversation about gravitational fields, high school rock bands, and the biggest data fusion project you can imagine.
FDL: What’s your background?
MR: For the most part, I’ve focused in Earth Sciences, specifically on the gravitational field of the Earth, radar data, optical data and remote sensing. In the last years however, I have been focusing on machine learning, deep learning and computer vision. I have found that by combining these two fields - Earth observation and machine learning - we can address some really interesting challenges.
FDL: Did you grow up wanting to be an Earth Science meets machine learning hybrid type person?
MR: Ha. Actually when I was growing up in Munich, Germany, i was a music kid. I was even in a rock band in high school with some friends.
FDL: Rock band in Germany… you mean like Rammstein?
MR: Not at all. More like brit rock. I played guitar. Now I really love to travel and at home I really like building my own computer hardware. I also have a passion for history and especially the Roman Empire. I listen to a lot of podcasts and spend a good deal of time thinking about how lessons from the past can give us perspective on the world we see today.
FDL: What has inspired you the most working as part of the FDL Europe team?
MR: Spending time with experts at the European Space Agency and UNICEF have really helped underscore for me that there is a much bigger challenge lying ahead.
The Earth is a complex environment. The world we live is driven by myriad systems - geophysical processes, weather meterology and on and on. Each system has spawned its own specific field of research. The BIG challenge today, and over the next decade, is to find ways to unify these individual silos of research into a common model for understanding our planet. This is about as big of a data fusion challenge as you can imagine. But if accomplished, the ramifications would be massive.
Today our FDL team is working on utilizing a few streams of data to make a difference in disaster response. But we are also looking at methods to connect many more streams of Earth observation data. With deep learning and machine learning we can start to see relationships across data sets and create much more complex physical models of the dynamics that shape our world.
About Marc Rußwurm
Marc recently started his PhD at the Technical University of Munich (TUM) following up on his Geodesy & Geoinformation studies. His scientific interest in Earth Observation was sparked at his Erasmus+ stay at the Earth Observation Group at the Polish Space Research Center. With this background, he joined the Computer Vision Research Group at the Chair of Remote Sensing Technology, TUM. Here Marc focuses on deep learning methods for multi-temporal image processing. He follows the perspective that the temporal dimension is as inherent to EO data as the spatial and spectral dimensions. Marc explored these ideas with recurrent neural network designs first in a study project and then in his Master Thesis. The results were presented at the CVPR Earthvision 2017 Workshop.