Computer Vision News - May 2022

5 VizWiz with Danna Gurari out how to design algorithms that analyze private information so that people don’t have to share it with a remote human . “ You need to understand each individual’s privacy concerns, ” she tells us. “ It’s an interesting aspect of this problem we’ve discovered through user studies with people who are blind. One privacy solution is not going tomeet all needs. It’s a niche problem of a set of larger problems we deal with. I think that inherently it’s often a few-shot learning problem. We have a new few-shot learning challenge this year, and we’ll see more in future years. It’s not that people have to focus specifically on privacy; it’s just focusing on the challenges that are intrinsically intertwined with privacy. ” big industry players who are keen to use them to help shape their products to meet customers’ needs, including Facebook , Google , Microsoft , Amazon , and IBM . Is this the dream scenario of scientific work having a practical impact on real people in the real world? “ That’s the hope! ” Danna asserts. “ Another goal for us is ensuring researchers are launched from academia into industry with the right skills to continue the progress of developing these technologies. It’s about helping to launch a whole new generation of developers . Every person that participates makes a difference. ” One surprising aspect of this work has been how much private information is included in the reviewed data, from credit cards to passports and even pregnancy tests. People are deciding the trade- off to share and learn about private information is more important than not sharing it . This discovery has created a newbranch of research for Danna, figuring Challenges for VizWiz Workshop “People are deciding the trade- off to share and learn about private information is more important than not sharing it!” Challenges for VizWiz Workshop

RkJQdWJsaXNoZXIy NTc3NzU=