Computer Vision News - November 2016

ECCV Daily: What is particularly challenging in doing that? Emma: We have two main parts of the project. We have a theoretical side where we prove a theorem that our constraint holds and that it’s the only constraint of its kind. To do that, we had to really delve into some deep mathematics. We had to use the theory of distributions rather than function. We had to learn a lot to handle all of the possible cases in a single framework. Then on the experiment side, to build a camera system that works takes a lot of time in the lab. We messed around with optics and found that there was a gap between the blur that we were expecting and the blur that we measured. We had to find solutions to those day to day problems.. ECCV Daily: How did you manage to solve that? Emma: Actually, we realized that we didn’t need to solve a lot of these problems. Instead of refining the optics, we get approximate solutions to our equation. We show experimentally that it’s good enough. We do this LeadSquared system. We developed a calibration procedure that optimized our depth measurement, and found that approximate solutions to our constraint are sufficiently robust to provide useful scene information, so that we could use a simple camera made of off-the-shelf, imperfect parts, rather than requiring highly precise, specialized optics. ECCV Daily: Who works with you? Emma: I work with another graduate student named Qi Guo . He’s great. He handled most of the experimental side. Then we worked with three professors; Sanjeev Koppal at the University of Florida, who has a lot of experience in micro-sensing, Steven Gortler at Harvard who helped out a lot on the math, and my advisor, Todd Zickler, who is so important to the whole project. ECCV Daily: What is the main teaching you learned from Todd Zickler? Emma: I think in terms of scientific work, he really cuts through abstractions and makes sure that we understand everything at a very basic level. For him, anytime you make a statement, he may ask you to go to the board, show the equation, and show how you got there. There’s no fear in asking simple questions, and there’s no shame in explaining simple things. That has really deepened my understanding and helped me communicate my understanding to other people. Our Pick at ECCV 13 BEST OF ECCV