Fraunhofer Institute for Industrial Mathematics ITWMAre acting and mathematics at odds? Dr. Henrike Stephani, deputy head of the department »Image Processing« at the Fraunhofer Institute for Industrial Mathematics ITWM, has tried both. In the end, she chose the scientific path – for a variety of reasons.
AI Instead of 007: The Exciting Cases Await in Mathematics
This Article Is Part of Our Book »Forscherinnen im Fokus – Wir schaffen Veränderung« (»Female Researchers in Focus – We Create Change«)
Studying theater studies in Berlin was intended to bridge the gap until a place at a state drama school became available. To get to know the profession better, Henrike did an internship as a director at the renowned Deutsches Theater in Berlin. »Somehow I didn't like the reality of the profession,« she looks back today. »I was also put off by the high probability that I wouldn't get a job that would finance my life.»So she followed a quote from actress Judi Dench, who flickered across the screen for many years as James Bond's boss »M«: »If you want to be an actress or can imagine doing something else, then do the other thing!«
Creative and Close to People With Mathematics
The need to learn particularly difficult things and the ever-present enthusiasm for the subject led Henrike to study mathematics. Not without respect: »People always say that mathematics is difficult. But the real challenge is not to get discouraged if you don't understand abstract things straight away. What I liked about my studies was that you can ›get stuck in‹,« she says. What's more, mathematics is not just a dry series of rules: »I'm one of the head people, the theorists. A lot happens on a meta-level, many trains of thought are enormously creative. Like in a good novel – there are surprising twists and turns that you would never have thought of, but also steps that happen one after the other in a comprehensible way.« Another thing that has never changed is that she enjoys interacting with lots of people: »That's why a career purely at university wouldn't have been for me. That left me with a job in management consulting or applied research.«
Henrike applied for a position at the Fraunhofer ITWM in Kaiserslautern. Her path here began in 2008 with her doctorate, led to a permanent position in 2012 and in 2014 to the position of deputy head of the »Image Processing« department. But for Henrike, the Fraunhofer ITWM is more than just a job. She is driven by helping to shape and change things. That's why she was active as an equal opportunities officer for many years, gets involved in strategic issues and is the contact person for experts and users of Artificial Intelligence in Rhineland-Palatinate as a consultant for the »AI pilot« Prof. Dr. Anita Schöbel.
AI Is Changing Work
So a lot has changed since 2008. Instead of classically modeled solutions, today it is AI with neural networks that takes on many tasks, presents researchers with challenges and demands new approaches. »Image Processing is one of the first disciplines in which it became particularly clear that AI can deliver better results than humans-at least if you are very precise at the beginning and end of the process.« A lot has also changed personally. Instead of active science, Henrike's everyday life is now dominated by management tasks. In addition to project management, applications and meetings, these also include business trips. In her absence, her husband, also a scientist at the Fraunhofer ITWM, looks after their two children. He has reduced his job to 90 percent, while she works full-time. »At Fraunhofer, we can also organize our working and home office hours very flexibly. I usually supplement an intensive work phase in the office until 3 p.m. with flexible work phases in the afternoon and evening at home, alternating with family time.«
Henrike: »Follow your talents, look out into the world with an open mind. If you are looking for something that you can use to make a difference and at the same time want interesting, committed people as colleagues, applied research is definitely something for you!«
Research With an Impact
But how do young people make the right decision regarding their future career path? Henrike says: »Follow your talents, look out into the world with an open mind. If you are looking for something where you can make a difference and at the same time want to work with interesting, committed people, then applied research is definitely for you!«
A Case for Deep Learning: Disaster Relief With Drone Images
Earthquakes, floods, fires: countless disasters turn countries into crisis zones and people into people in need of help. Scientists like Henrike have developed the »EDDA« software using AI-supported image analysis. It improves emergency coordination and thus speeds up important relief operations.
For a long time, satellite images were the only way to assess the situation in disaster areas and plan the deployment of aid workers. However, the process of evaluating the images took too long and sometimes led to misinterpretations that resulted in misdirected aid supplies.
Always in the Right Place With »EDDA«
New technological possibilities such as unmanned drones have already made the process faster thanks to real-time aerial images. However, the time that has just been gained melts away when viewing the many individual images. With the »EDDA« software, researchers like Henrike and her colleagues have solved this problem too. The head of the project is her department head Markus Rauhut. They have developed Image Processing and deep learning algorithms that can be used to evaluate drone images more quickly and accurately. Users only need a laptop, not even an internet connection is required.
Software That Learns on Its Own
This ultimately simple process requires a great deal of preparatory work on the part of the scientists. With considerable manual effort, they first add background information to data such as satellite images from previous relief operations and model data: What is the condition of the road, what is the condition of the houses, can the bridges still be used and how many people live in the respective location? Information on water levels can now also be displayed. These images are fed into the software's neural network, which is modeled on the human brain. »EDDA« then goes into training in real crisis areas. Here, it learns to independently capture, classify and analyze newly captured photos based on the stored data and Image Processing and deep learning algorithms. Because there is always new data with new information and this information is constantly being linked to one another, »EDDA« automatically improves from mission to mission.
Frau N. Hofer searches the disaster area with a drone.
Working Together for Humanitarian Aid
The »EDDA« project was funded by the Fraunhofer Foundation. This is another reason why NGOs and emergency teams can use the application free of charge. The researchers are also working closely with initiatives such as the United Nations World Food Program (WFP). They provide the important real data from crisis areas, enable practical training and, with their feedback from real operations, help to ensure that »EDDA« can continue to provide valuable help in emergencies in the future.