CVPR Daily - Wednesday

24 DAILY CVPR Wednesday Women in Computer Vision interesting, especially when students had questions that I didn’t have the answer to at that point. So we had discussions. Let’s say you have to choose between a career in a lab doing pure research or a career in academia creating a new generation of successful students and scientists. I don’t know! I think it’s still too early to tell. I think right now I’m more interested in the research path. I don’t think I’ve had enough experience managing or mentoring students yet. Tell us more about your research. My previous works have been exploring how to codify real-world scans. Real world scans are very noisy and broken. Basically, it’s very incomplete geometry. You want to replace it with an artist-designed, high quality model. How do we do that? My previous works have revolved around this problem. One good way to do this is to retrieve a shape from some existing repository of artist generated models, then deform it to fit the given target. What we’re saying is that the retrieval process should not be just by geometric similarity, but you should retrieve based on the formability: when you don’t retrieve something that is similar in terms of their shape and geometry, but something where you can retrieve it and throw it into a deformation function, such that it will deform to the given target. What’s your message for the CVPR community? I hope we all get to see each other in person at some point post-pandemic! Next year in New Orleans! Read more than 100 interviews with Women in Computer Vision!