Thomas Pöllabauer

Who are you and what do you do? Who are your partners?
My name is Thomas Pöllabauer. After studying computer science and visual computing at TU Berlin and TU Darmstadt respectively, I am currently working as a research assistant at the Fraunhofer Institute for Computer Graphics Research. I work on machine learning, especially deep learning, in the computer vision context. This is also the field of my Software Campus project (short name: Differentiable Federated Pose). I’m working on it with the Fraunhofer Gesellschaft as a research partner and Software AG as an industry partner.

What do you expect from the Software Campus?
Computer science permeates our world and is, in my eyes, one of the most interdisciplinary working environments. That makes goal-oriented planning and prudent management of a project all the more important, both competences that I would like to deepen in the course of the SWC.

What is the content of your IT project and how could it be applied in the future?
The dominant part of Deep Learning methods that can be used in productive systems is based on the premise that large amounts of data are available to train a model and that these data can be provided with high-quality labels. In fact, there are a large number of interesting problems where no primary data exists, for example because it poses a security risk. Think of a diagnostic imaging procedure in a hospital. The release of this data may conflict with patients’ right to privacy. One approach to protect this is to adapt federated learning, whereby the data is processed exclusively on site and only parameters derived from it are accumulated centrally. However, if the data does not leave the recording site, then no external entity can label it – labels that are, however, necessary for most conventional procedures. My project aims to use Differentiable Rendering to obtain labels of sufficient quality to be able to train the system completely on site and without human intervention. The application here is object pose recognition, whereby no real images are available until the time of the first application. Ultimately, however, the method can be applied to a wide range of vision problems.

What fascinates you about computer science?
Computer science is full of exciting tasks and you can start dealing with the subject matter with any level of prior knowledge, see here.

Which app/technical invention is indispensable for you?
Electricity. Without electricity, there’s no coffee in the morning, and as you learn in the introductory course of computer science studies: a computer scientist turns coffee into code (and possibly paper).

Has one personality from computer science or management particularly impressed you?
Elon Musk. You may be more or less taken with his human side, but I find it hard not to be impressed by what he does in his various companies. Some people dream of Mars. This man doesn’t just dream, he builds the rocket! Granted, he’s a little behind schedule, but we know how long it can take to open an airport …

In your opinion, is there “the” quality that a top manager has to have today in order to be successful?
Under the premise that successful = earns a lot of money doesn’t mean true or top, I would say: you are successful if you earn the respect of employees, partners and customers.

Imagine: Internet breakdown for 1 month – what do you do?
Stay calm, I don’t get bored without the internet. We’ve already discussed this above: give me electricity and coffee and I’m good to go. A notebook would still be nice, even if it’s made of paper and with a pencil. Incidentally, I see a good opportunity to seize a potential disaster (no x from {Netflix, Youtube,Facebook…} available) as an opportunity. You shouldn’t underestimate what you can do in a bit of uninterrupted time.