学术报告:Visual concept search learned from social media

题目: Visual concept search learned from social media

时间:2012年6月20日(周三) 上午9:30 – 11:00

地点:北京大学计算机所大楼106报告厅

Abstract:

In a world with increasing amounts of digital pictures, content-based image retrieval is an important and scientifically challenging problem in ICT research. In this talk, I will present my work on tackling the problem by learning many visual concepts from social media. While social image tagging is generating massive amount of labeled examples, social tags are known to be subjective and noisy. As a consequence, directly learning visual concepts from socially tagged examples is problematic. I will introduce algorithms we developed to harvest high-quality training examples from the social web, with no need of extra manual labeling. I will demonstrate the effectiveness of our algorithms for several tasks including social image retrieval, visual categorization, and complex visual searches in unlabeled data.

Speaker:

Dr. Xirong Li is currently an assistant professor at the MOE Key Lab of Data Engineering and Knowledge Engineering, Renmin University of China. His research is on visual search and multimedia content analysis.

He received the Ph.D. degree from the University of Amsterdam (2012), the Master and Bachelor degrees from the Tsinghua University (2007 and 2005), all in computer science. He received the Best Paper Award of the ACM International Conference on Image and Video Retrieval 2010, Best Paper Nominee at the ACM International Conference on Multimedia Retrieval 2012, and the Chinese Government Award for Outstanding Self-Financed Students Abroad 2011. He has published a number of papers in top-tier journals and conferences including IEEE Transactions on Multimedia and ACM Multimedia, with 600+ citations and an H-Index of 12.

CLOSE

上一篇 下一篇