好文作者面授招–20151112

【15-34期VALSE Webinar活动】

报告嘉宾:庄连生(中国科学技术大学)
报告时间:2015年11月12日(星期四)晚20:00(北京时间)
报告题目: Sparse Illumination Learning and Transfer for Single-Sample Face Recognition (稀疏光照迁移技术在单样本人脸识别中的应用) [Slides]
主持人:王琦(西北工业大学)
报告摘要:
Single-sample face recognition is one of the most challenging problems in face recognition. We propose a novel algorithm to address this problem based on a sparse representation based classification (SRC) framework. The new algorithm is robust to image misalignment and pixel corruption, and is able to reduce required gallery images to one sample per class. To compensate for the missing illumination information traditionally provided by multiple gallery images, a sparse illumination learning and transfer (SILT) technique is introduced. The illumination in SILT is learned by fitting illumination examples of auxiliary face images from one or more additional subjects with a sparsely-used illumination dictionary. By enforcing a sparse representation of the query image in the illumination dictionary, the SILT can effectively recover and transfer the illumination and pose information from the alignment stage to the recognition stage. Our extensive experiments have demonstrated that the new algorithms significantly outperform the state of the art in the single-sample regime and with less restrictions. In particular, the single-sample face alignment accuracy is comparable to that of the well-known Deformable SRC algorithm using multiple gallery images per class. Furthermore, the face recognition accuracy exceeds those of the SRC and Extended SRC algorithms using hand labeled alignment initialization.
参考文献:
[1] Single-Sample Face Recognition with Image Corruption and Misalignment via Sparse Illumination Transfer. Liansheng Zhuang, Allen Yang, S. Shankar Sastry, Zihan Zhou, Yi Ma. In Proceedings of CVPR 2013, 2013.
[2] Sparse Illumination Learning and Transfer for Single-Sample Face Recognition with Image Corruption and Misalignment. Liansheng Zhuang, Tsung-Han Chan, Allen Y. Yang, S. Shankar Sastry, Yi Ma. Accepted by International Journal of Computer Vision (IJCV), July 2014.
报告人简介:
庄连生,博士,中国科学技术大学信息学院讲师,中国计算机学会多媒体专委会副秘书长、中国图像图形学会多媒体专委会副秘书长。分别于2001年(本科)、2006年(博士)毕业于中国科学技术大学,2006年留校工作至今。2011.3-2011.11入选微软亚洲研究院“铸星计划”(合作者:马毅博士),2012-2013在美国加州大学伯克利分校做访问学者。庄连生博士的主要研究领域是计算机视觉和机器学习,尤其关注物体识别和半监督学习。相关研究成果已发表在IJCV, T-IP,CVPR等国际期刊和国际会议上,并获得了HHME2014最佳论文奖。先后主持了多项国家自然科学基金(包括青年项目、面上项目、重点项目子课题)、863项目的研究。他是IEEE会员、ACM会员,并担任IEEE T-IP、IEEE T-MM、CVPR 2014、CVPR2015、ICCV2015等多个国际期刊和会议的审稿人。

报告嘉宾:林国省(The University of Adelaide)
报告时间:2015年11月12日(星期四)晚21:00(北京时间)
报告题目:Learning Deep Structured Models for Semantic Segmentation. [Slides]
主持人:彭玺(A*STAR)
报告摘要:
The fist part of the talk is about how to explore the context by learning deep structured model. We achieve an intersection-over-union score of 77.8 on the challenging PASCAL VOC 2012 dataset, which is a new record.
The second part of the talk concerns a new deep structured learning method. We propose to directly learn the CNN based message estimator in message passing inference, instead of learning conventional potential functions.
参考文献:
[1.] Gushing Lin, Cfhunhua Shen, Ian Reid, Anton van dan Henge; Efficient Piecewise Training of Deep Structured Models for Semantic Segmentation; arXiv.
[2.] Gushing Lin, Cfhunhua Shen, Ian Reid, Anton van dan Hengel;Deeply Learning the Messages in Message Passing Inference; NIPS 2015.
报告人简介:
林国省现任澳大利亚阿德莱德大学博士后研究员。2014年获阿德莱德大学博士学位,师从沈春华教授。其研究兴趣包括 Structured Learning,Deep Learning, Image retrieval, and Semantic Image Segmentation。博士期间获得Google PhD Fellowship (one of 38 winners world-wide in 2014). 目前已发表2篇 TPAMI, 2篇NIPS/ICML和4篇CVPR/ICCV/ECCV。

(Visited 693 times, 1 visits today)