Southwest Jiaotong University School of Mathematics


ag网站亚游登录  >  学术科研  >  信息与计算科学系  >  正文


ag网站亚游登录:   作者:代数编码及其应用团队     日期:2019-05-22 21:58:53   点击数:  




报告地点: X7503


Title: Time-evolving graphs for anomalous event identification and prediction

报告摘要:A novel architecture is proposed to address anomalous event detection and prediction task as a graph stream analysis task. In particular, object movement information is extracted using an optical flow based scheme and represented as dynamic graphs, where individuals are denoted as nodes, and the behavior patterns, which are represented using similarity measurements, are used as edge weights. Then we propose a max-flow/min-cut surveillance (MFMCS) scheme to detect the component(s) that would represent the bottleneck of the network, which are eventually identified as anomalous events or conducting crowd event identificaton.

报告人介绍:Meng Yang is a PhD graduate from the University of Melbourne, the Department of Computing and Information Systems. Her research interests contain data mining, machine learning, deep learning and computer vision. During her PhD candidature, she has published papers in IJCNN, Pattern Recognition, etc. She was also awarded Google PhD travel scholarship.


XML 地图 | Sitemap 地图