Event Location: NSH 1507
Bio: Prof. Fei-Fei Li’s main research interest is in vision, particularly high-level visual recognition. In computer vision, Fei-Fei?s interests span from object and natural scene categorization to human activity categorizations in both videos and still images. In human vision, she has studied the interaction of attention and natural scene and object recognition, and decoding the human brain fMRI activities involved in natural scene categorization by using pattern recognition algorithms. Fei-Fei graduated from Princeton University in 1999 with a physics degree. She received PhD in electrical engineering from the California Institute of Technology in 2005. From 2005 to August 2009, Fei-Fei was an assistant professor in the Electrical and Computer Engineering Department at University of Illinois Urbana-Champaign and Computer Science Department at Princeton University, respectively. She is currently an Assistant Professor in the Computer Science Department at Stanford University. Fei-Fei is a recipient of a Microsoft Research New Faculty award and an NSF CAREER award. (Fei-Fei publishes using the name L. Fei-Fei.)

Abstract: The world is awash with data. But as academic researchers, we never have enough. The field of computer vision has mostly been using Caltech101/256 and PASCAL VOC datasets to benchmark and evaluate recognition algorithms, at least 3-4 orders of magnitude smaller than what humans can do. To bridge this gap, we have recently put together a new image dataset called ImageNet (www.image-net.org), currently consisted of more than 10 million images across 15,000+ visual categories, all collected from the web and verified by humans. The construction of ImageNet has been a tremendously challenging process, forcing us to dive into relatively unchartered water of crowdsourcing technology. Using ImageNet as a resource, we present here a series of unpublished work in benchmarking existing algorithms for image classification task by using more than 10,000 visual concepts. By exposing ourselves to such large-scale, we observe issues and formulate new insights towards real-world scale recognition problems that might not be seen in smaller scale experiments. I will particularly talk about four areas: computation, scale, data density, and concept hierarchy. For the remaining of the talk, I will share with you a number of interesting projects that is ongoing in the ImageNet project. Last but not least, stay tuned for the upcoming ImageNet Challenge 2010, in partnership with with PASCAL VOC.