ECCV 2016 Daily - Thursday

ECCV Daily: Ranjay, can you tell us about working with Fei-Fei? Ranjay: She’s good at a lot of different things. She gives you amazing advice, not just on the high level concepts and which problems to tackle, but she also digs in and gives you a lot of low level advice on how to formulate your papers, how to formulate your talks, or how to actually tackle a problem from the right direction. For example, there are a lot of ways you could have tackled the problem that I worked on at ECCV. She’s very good at identifying what hasn’t been done before. She has a breadth of knowledge of all of the things that have been done. ECCV Daily: Is she very demanding? Ranjay: She does a good job of pushing a student knowing that the student can deliver on those particular points. She knows everyone’s potential so she treats every student differently. ECCV Daily: If there was one skill you could have of Fei-Fei’s, which one would it be? Ranjay: The ability to multi-task - I think a lot people take for granted how hard it is to manage all of the projects going on in the lab, apply for grants, give a Ted Talk, have children, spend time with family, have a life, and maybe do something outside of research in this industry. Balancing is something that I haven’t gotten to yet and something I would love to have. ECCV Daily: Tell us about your work. Ranjay: Computer vision has focused a lot on detection of objects; looking at an image an understanding all of the different things that they see. What my work is doing is pushing that boundary to, not only detect the objects, but to be able to understand how they are related to each other. It’s one thing to be able to say that I see a person and a llama. It’s another thing to say that the llama is chasing the person vs. the llama is gently nudging the person. The different contexts will help us build smarter agents that can reason about which actions to take even if they see the same objects. ECCV Daily: What was the most difficult part for you in making this Visual Relationship Detection with Language Priors 10 ECCV Daily : Thursday Presentations Ranjay Krishna is a PhD student at Stanford, working mainly with Fei-Fei Li