CVPR Daily - Thursday

Many in our community know you well, Anna. But some people are at CVPR for the first time: can you share with the newcomers what you do? This is a great point. I discovered when preparing for CVPR that many junior students have never done a poster session, and they don't know how giant these things are and how giant the posters are. I felt compelled to inform everyone that you have to put high-quality graphics and very clear and well-structured posters so that your life is easier. Speak loudly and drink plenty of water. I remember my very first poster session. Oh my God! I definitely lost my voice then. I think this is fun. This is fun for the first- timers. We should make a special CVPR only for first-timers. Maybe one day I will tell you about my first CVPR. Tell me, Anna, what is your work about? I am working at the intersection of vision and language. This bridges many different directions from the more traditional vision language tasks. I started working in the early days of captioning and visual question answering. More recently, the horizons have expanded to what else language can do for vision. With the advancements of recent large- scale models, many of them multimodel models, open new possibilities for how we can expand the zero-shot capabilities of our model's generalization. Or, even use language to advise models like humans, about what to do and what not to do. I'm very passionate about this direction of leveraging language for vision. This convergence is starting to make a real impact. We've been hearing about it for a few years, but in the last one or two years, it's starting to become a real thing that many people are working on. Is that right? 19 DAILY CVPR Thursday Anna Rohrbach UK AINE CORNER Younger in Odessa