ECCV 2022 - Wednesday

scenes and the 3D world in an easy and controllable way. There are compact learned mathematical models underneath all these to do this. 3D has been an important part of the best years of your career. Can you tell us what you find so fascinating about it? Firstly, I’m not sure these are the best years of my career yet – hopefully, it will get better and better! [ he laughs ] We live in a 3D world, which we’re trying to fit into computers. Attempting to understand the 3D world around us without modeling the third dimension doesn’t seem like the right approach. If you model things in 3D, many of the invariances desired in algorithms are already built in, like rotational invariance – if you rotate an object, it doesn’t change its shape or appearance. Humans and animals learn to perceive and interact in the 3D world, so replicating that in machines just makes sense. Do you see the same passion in new teachers and students, or is it unique to you? We’ve seen a growth in interest in algorithms that incorporate 3D processing. We were more of a minority a few years back, but now there are as many papers doing 3D things as there are doing pure 2D things. What is themost precious learning that you have learned fromyour teachers? 11 Gerard Pons-Moll I learned different things from different people. One is to really focus on important problems and to think more about those problems than you do your next paper. I always try to go to the core of a problem. What’s the thing that will make the biggest difference? What’s the easiest way we can make that difference? What’s the simplest idea? If we publish a paper and people read it and think, oh, that’s obvious, but it produces excellent results, then we are successful because these are the key things we need to discover. Also, I’ve learned that helping your students think about their careers is very beneficial for them, but indirectly,