CVPR Daily - Wednesday

21 DAILY CVPR Wednesday parametric shifts. I like to understand how discrete and continuous data marry, and how to design your networks and machine learning models to better express these data. Why do we need that? The world that we live in is in 3D! The actual representation of data is 3D. Creating this 3D data in digital format requires very tedious and specialized processes by expert designers. In order for us to create more of these with less effort, we want to automate. These models are, by nature, discrete and continuous, like discrete shapes with continuous variation. We want to understand this data. What is so exciting in solving that problem? Problem solving in general is interesting for me. You have a list of properties that you want to satisfy, and how do you design your solution such that you still satisfy these given assumptions. This problem solving by itself is interesting. How I ended up in this area with these types of problems is largely because of my advisor. Leo Guibas guided me. I had a lot of help from postdocs and seniors in the labs through my initial research. The people who guided me through my research (apart from Leo of course) are seniors and postdocs from the lab, and mentors from my last internship in Adobe, who I actually published the CVPR paper with. That collaboration was also a lot of fun! Mikaela Angelina Uy