Computer Vision News - July 2022

BEST OF CVPR 18 Women in Computer Vision computational fluid dynamic solver and be able to say something about themovement and physics for application to design. What chance do you give yourself to achieve this? With a lot of collaboration, probably pretty high. The challenge is a lot of the animals that we study aren’t the classical opaque bodies. These are more transparent, translucent animals. There are really no comparable models out there to do that kind of work.  What do you miss the most, from aeronautics? Do you regret anything? I mean, if I can regret not being an astronaut, but I feel like I don’t have very much control over that process. I think what I miss themost is being able to do very straightforward repeatable measurements or studies of physics, where once you start delving into biology and organism systems, there’s almost no replicability. It is so OK, I don’t need to introduce you. What would you like to achieve in this field? this field? That’s a good question. A couple. The reason why I’m here at CVPR this year is because we’ve been building out a label data set called FathomNet to help with fine grain categorization, to try to automate processing visual data with the long-term goal of being able to detect when we’ve come across animals that are known or unknown to science. So to me, having a data pipeline in place where we can know right away if there’s a new animal completely undescribed that needs further study. Being able to do that whole process would be fantastic. For the bio- inspired design stuff, what I would love to be able to do is kind of multi-view, three-dimensional full reconstructions of animals that are time-varying so that you can take that reconstruction and put it in a MBARI