Computer Vision News - November 2022

24 Women in Computer Vision set every time you have a new problem. Your message for the community? I think this community is pretty awesome to be in right now. So many things are changing, and that's really exciting! Lots of new ideas and new tools. Do you want to give one example of something that’s impressing you at this moment? Thework that people have done leveraging diffusion models in order to provide supervision for 3D is pretty cool. So, for instance, DreamFusion, I think it’s called, actually showed that you can get some notion of 3D structures out of image- based models. I don't think it will solve everything, but I think that it was pretty cool. More than 100 inspiring interviews with successful Women in Computer Vision! The big thing that I think would be cool is to take a short cell phone video of an environment and then import it into a videogame level because I like toplayvideo games. [ laughs ] I think the video game sort of nature of things, the interactivity is really cool, and that's something that we don't really have remotely solved yet. So you're that kind of millennial that likes video games? Actually, I started playing video games in grad school, so I'm a bit late. [ laughs ] Angela, tell me one thing about you that we don't know. Besides the fact that I like to play video games, I'm trying to jog everywhere I go to a conference. So I went jogging along the beach in Tel Aviv this morning which was really nice. It’s a bit warm for jogging, but it's super pretty. I used to roller skate there. But then I broke too many bones to continue. I hope that your jogging is less brutal! No broken bones yet! [ laughs ] Let's talk about future projects… maybe one word about things that you have not done yet. I think a lot of the stuff that we're looking at involves how we learn more efficiently. So, in 3D, we don't really have anywhere near as much data as you have with images or with text, which means that you do have to think about a more efficient learning paradigm and how we can learn with fewer labels or no labels. Or create the data? Or create thedata!  That is always anoption to jumpstart a new area. It's obviously not a tractable solution to create a new data BEST OF ECCV