Computer Vision News - November 2016

than the speed of sound is a sort of canonical example. We’ve retreated from that. We don’t fly faster than the speed of sound. We’re not pushing toward the speed of light. Rather than pushing those physical limits, we’re pushing more technological and communication limits. Well, maybe that makes sense. Humans want to communicate so this is where we put a tremendous amount of energy. Ralph: On Monday, during your presentation, you rephrased someone’s question as “Am I really right to work on optical flow? This is a question that I’ve asked myself for 30 years.” Do you have an answer to give to yourself? Michael: As scientists, we have a luxury of pursuing problems that we think are interesting and may prove important. I had an experience during my PhD. You write your thesis. You have an introduction and a conclusion. Then you want to say why you are doing this. I was doing my research on optical flow at the time. I wrote a bunch of things about why I thought optical flow was important. I thought it would be important for robotics to find occlusion boundaries or surface boundaries, to understand scene and its geometry, and so on. I really believed that. Finally, we couldn’t do optical flow fast enough for it to be useful for a robot, but I imagined that would be the big use for it. Then I made my code available online, which wasn’t so easy in 1992. People downloaded my code and used it. It was very humbling because people took this code that I wrote, and I thought I knew what they would use it for, but they used it for everything else, as I could never imagine. People in biology were using it to track cells. It had never occurred to me that you could use it to track cells. People used it in Hollywood, so it got used in making the movie “ What Dreams May Come ”, where they created this painterly effect. They used it for computer graphics. It got used in “ The Matrix Reloaded ” to do facial animation. Things that I never would have dreamed of in my life. Ralph: Why was that humbling? Michael: Because you think you might know why you are doing something. What’s a tool? A new tool gives people ability to manipulate something in a way that they could never do before. The world has problems. I have no idea what those problems might be. People are looking for tools to solve their problems. When a new tool comes out, they will adapt it, and solve problems in all kinds of ways. It was humbling in the sense that, after that experience, I thought it is naive of me to assume I know how my technology will be used. I can’t “People took this code that I wrote, but they used it for everything else, as I could never imagine” 18 Interview with Michael Black BEST OF ECCV