20160504-13 马超: Hierarchical Convolutional Features for Visual Tracking

2016-4-29 21:01| 发布者: 程一-计算所| 查看: 10092| 评论: 0

摘要: 报告嘉宾1：马超（上海交通大学）报告时间：2016年5月4日（星期三）晚20:00（北京时间）报告题目：Hierarchical Convolutional Features for Visual Tracking主持人：吴毅（南京信息工程大学）报告摘要：Visual obj ...

报告嘉宾1：马超（上海交通大学）

报告时间：2016年5月4日（星期三）晚20:00（北京时间）

报告题目：Hierarchical Convolutional Features for Visual Tracking

主持人：吴毅（南京信息工程大学）

报告摘要：Visual object tracking is challenging as target objects often undergo significant appearance changes. In this talk, I will present our recent work on how to best exploit features extracted from deep convolutional neural networks trained on object recognition datasets to improve tracking accuracy and robustness. The outputs of the last convolutional layers encode the semantic information of targets and such representations are robust to significant appearance variations. However, their spatial resolution is too coarse to precisely localize targets. In contrast, earlier convolutional layers provide more precise localization but are less invariant to appearance changes. We interpret the hierarchies of convolutional layers as a nonlinear counterpart of an image pyramid representation and exploit these multiple levels of abstraction for visual tracking. Specifically, we adaptively learn correlation filters on each convolutional layer to encode the target appearance. We hierarchically infer the maximum response of each layer to locate targets. Extensive experimental results on a largescale benchmark dataset show that the proposed algorithm performs favorably against state-of-the-art methods

参考文献：

[1] Hierarchical Convolutional Features for Visual Tracking. Chao Ma, Jia-Bin Huang, Xiaokang Yang, and Ming-Hsuan Yang. In ICCV, 2015

[2] Long-Term Correlation Tracking. Chao Ma, Xiaokang Yang, Chongyang Zhang, and Ming-Hsuan Yang. In CVPR, 2015

报告人简介：

Chao Ma is a PhD student in Shanghai Jiao Tong University, supervised by Professor Xiaokang Yang. He took a two-year study (2013.09-2015.09) in University of California at Merced and was working with Professor Ming-Husan Yang. He is interested in computer vision and machine learning problems. Typically, he is working on visual object tracking, single-image super-resolution, and sketch image retrieval (recognition). He has published 5 papers in the cutting-edge computer vision conferences and journals including ICCV/CVPR/ECCV/BMVC/IVC.

马超，上海交通大学博士研究生，导师杨小康教授，于2013年九月至2015年九月在加州大学默塞德分校Ming-Hsuan Yang教授的计算视觉组访问学习。研究兴趣主要关注在目标跟踪，图像超分辨以及图像检索。已在视觉领域的重要国际会议和期刊上发表5篇论文，包括ICCV，CVPR，ECCV，BMVC，和IVC。

报告材料[Slides]

收藏邀请

上一篇：20160420-12 吴祖煊: Video Content Recognition with Deep Learning下一篇：20160504-13 Shenlong Wang: 3D Scene Understanding through Maps of the World

20160504-13 马超: Hierarchical Convolutional Features for Visual Tracking

最新评论

相关分类