报告嘉宾: 罗烨(同济大学)
报告时间:2018年01月24日(星期三)晚上20:00(北京时间)
报告题目:
Modeling the Temporality of Visual Salilency and Its Application to Action Recognition
主持人:吴金建(西安电子科技大学)
报告摘要:
In this presentation, we mainly introduce the following two parts. At first, we investigate the temporality aspect of saliency estimation. A principled method based on three levels of saliency has been proposed: the intra-trajectory level, the inter-trajectory level and the static level. Experimental results validate the concepts put forth in the paper, as well as characterizing the effects of time, and the contributions made by individual levels. At last, as an extension of our proposed video saliency, we elicit from a fundamental definition of action low-level attributes that can reveal agency and intentionality. These descriptors are mainly trajectory-based, measuring sudden changes, temporal synchrony, and repetitiveness. The direct result of these descriptors is called the actionness map and it can be used to localize actions in a way that is generic across action and agent types. Experimental results on various datasets show the advantages of our method on action detection and action recognition comparing with other state-of-the-art methods.
报告相关文献列表:
1. Ye Luo, Loong-Fah Cheong and Tran Lam An, “Actionness-assisted Recognition of Actions”, in International Conference on Computer Vision (ICCV) 2015, pp. 3244-3252.
2. Ye Luo, Loong-Fah Cheong and John-John Cabibihan, “Model the Temporality of Saliency”, Asian Conference on Computer Vision (ACCV) 2014, Vol. 9005, pp. 205-220.
3. Ye Luo, Junsong Yuan, Ping Xue and Qi Tian, “Saliency Density Maximization for Efficient Visual Objects Discovery”, in IEEE Trans. on Circuits and System for Video Technology (TCSVT), Vol. 21, pp. 1822-1834, 2011.
报告人简介:
罗烨博士现为同济大学软件学院助理教授。2010年在新加坡南洋理工大学取得博士学位,同年在新加坡国立大学进行博士后的研究工作。2016年7月份,加入同济大学。她以第一作者及通讯作者身份,发表国际期刊和国际会议20余篇。参与国家自然科学基金面上项目4项,参与新加坡科研类项目4项,承担中央高校基本科研业务费专项资助项目1项。2015年获得国际计算机视觉顶级会议ICCV 优秀青年研究者奖。承担Computer Vision and Image Understanding,the Visual Computer,Signal Processing: Image Communication,IEEE Signal Processing Letter,Neurocomputing,ICIP2017, ICME 2015, ICME 2014, ACCV 2010 等多个国际期刊和国际会议的审稿人。
讲者个人主页:
http://sse.tongji.edu.cn/Data/View/3143
特别鸣谢本次Webinar主要组织者:
VOOC责任委员:吴金建(西安电子科技大学)
VODB协调理事:王琦(西北工业大学)
活动参与方式:
1、VALSE Webinar活动依托在线直播平台进行,活动时讲者会上传PPT或共享屏幕,听众可以看到Slides,听到讲者的语音,并通过聊天功能与讲者交互;
2、为参加活动,请关注VALSE微信公众号:valse_wechat 或加入VALSE QQ群(目前A、B、C、D、E、F群已满,除讲者等嘉宾外,只能申请加入VALSE G群,群号:669280237),直播链接会在报告当天(每周三)在VALSE微信公众号和VALSE QQ群发布;
*注:申请加入VALSE QQ群时需验证姓名、单位和身份,缺一不可。入群后,请实名,姓名身份单位。身份:学校及科研单位人员T;企业研发I;博士D;硕士M。
3、在活动开始前10分钟左右,讲者会开启直播,听众点击直播链接即可参加活动,支持安装Windows系统的电脑、MAC电脑、手机等设备;
4、活动过程中,请勿送花、打赏等,也不要说无关话语,以免影响活动正常进行;
5、活动过程中,如出现听不到或看不到视频等问题,建议退出再重新进入,一般都能解决问题;
6、建议务必在速度较快的网络上参加活动,优先采用有线网络连接;
7、VALSE微信公众号会在每周一推送上一周Webinar报告的总结及视频(经讲者允许后),每周四发布下一周Webinar报告的通知。