VALSE

VALSE 首页 活动通知 好文作者面授招 查看内容

20160803-25 江焯林:Visual Recognition via Submodularity and Sparse Representatio ...

2016-7-30 13:13| 发布者: 程一-计算所| 查看: 5830| 评论: 0

摘要: 报告嘉宾1:江焯林(Raytheon BBN Technologies, USA)报告时间:2016年8月3日(星期三)晚20:00(北京时间)报告题目:Visual Recognition via Submodularity and Sparse Representation主持人:王瑞平(中科院计算所) ...

报告嘉宾1江焯林(Raytheon BBN Technologies, USA)

报告时间:20168月3日(星期三)晚20:00(北京时间)

报告题目:Visual Recognition via Submodularity and Sparse Representation 

主持人:王瑞平(中科院计算所)


报告摘要:

Feature extraction is critical for the final performance of computer vision recognition algorithms, however, most of current prevalent features are hand-crafted, such as SIFT, HOG and LBP etc. We have developed representation learning techniques based on sparse representation and submodular functions, which can automatically learn discriminative representations from visual data. In this talk, I will firstly present a submodular learning framework for discriminative attribute selection for action recognition. Our selection framework is very general and able to achieve state-of-the-art performances. Then I will present a novel sparse representation based feature learning approach called Label Consistent K-SVD (LC-KSVD). This approach and its online learning version have been successfully applied to many other applications. Finally I will introduce a new supervised deep neural network called Label consistent Neural Network (LCNN). The class-specific representations generated by late hidden layers are discriminative enough to achieve good performance using a simple kNN classifier. The representations generated by LCNN can be directly used in other applications.


报告相关文献列表:

Paper title, authors, Journal, 2015

Label Consistent K-SVD: Learning A Discriminative Dictionary for Recognition. TPAMI, 2013

Submodular Attribute Selection for Action Recognition in Video. NIPS, 2014

Learning Discriminative Features via Label Consistent Neural Network. arXiv:1602.01168, 2016


讲者简介:

江焯林,美国Raytheon BBN Technologies研究科学家(Research Scientist)2013-2014年,华为(香港)诺亚方舟实验室研究员。2010-2012年,美国马里兰大学高级计算机研究所研究科学家。2012年,所著论文“discriminative dictionary learning with pairwise constraints”获得亚洲计算机视觉学术会议(ACCV 2012)最佳学生论文奖。担任模式识别领域国际顶级期刊《Pattern Recognitionspecial issue客座编辑, CVPRICCVECCVIJCAIACMLACCVEurographics等重要国际会议的程序委员会委员,《TPAMI》、《IJCV》、《TIP》、《TMM》、《TNNLS》等重要期刊审稿人。已发表文章30多篇,其中发表在国际顶级会议和期刊(包括CVPRICCVNIPSTPAMITIP)的文章有14篇,被引用次数已经超过1500次。申请中国发明专利13项,美国发明专利2项。


最新评论

小黑屋|手机版|Archiver|Vision And Learning SEminar

GMT+8, 2021-1-25 22:53 , Processed in 0.013303 second(s), 15 queries .

Powered by Discuz! X3.4

Copyright © 2001-2020, Tencent Cloud.

返回顶部