Computer Vision News - January 2021

321 Siyu Tang straightforward. A direct application would be autonomous driving. We would want to understand people's motion and location from images and videos. Our recent work is egocentric activity recognition. It is more related to the AR and VR scenario. We want to understand behaviors in this type of scenario. For one recent work about modeling hand-object interaction, a direct application would be to generate synthetic training data. That’s also an application of the model we are developing. Did anybody influence you in this direction? Yes! Yes! I have two amazing mentors and advisors. I think I’m a super lucky person. My first is my PhD advisor, Bernt Schiele. He is a really, really supportive and amazing mentor. I knew nothing about research. I knew a little bit about computer vision. He taught me from the very beginning how to do research, how to really aim for good, highly impactful scientific work. That I learned from Bernt. My postdoc advisor, Michael Black, is also super amazing. I really learned a lot from him, how to find super challenging scientific questions and how to really tackle these scientific questions. I learned a lot from both. I agree: you are very, very lucky! [ laughs ] Yes! What was the most difficult thing for Bernt to teach you? He once commented on how I can do things amazingly fast. He thinks it’s really a talent, but it also comes with a price. Sometimes I don’t really care about details. Details are important. Yes, details are super important. Sometimes details really matter, whether it works or not. Bernt really taught me how to have both: working really fast to progress and at the same time, to think about all of the details. He spent time teaching me these kinds of things. “I can really concentrate and be productive if my goal is very clear”