南京大学计算机科学与技术系

软件新技术与产业化协同创新中心

摘 要：

With the advent of social media, watching and filtering of images posted on social media is of current research interest. People post images and videos on social media to express their feelings or emotions according to their behaviors and this type of social media data is increasing day-by-day. This work focuses on classifying person behavior-oriented social media images, namely, bullying, threatening, depressed, sarcastic and psychopathic along with extraversion (normal). The proposed approach first detects faces as a foreground component, and other information (non-face) as background components to extract context features. Next, for each foreground and background component, we explore the Hanman transform to study local variations in the components. Based on Hanman (H) values, the proposed approach combines the H values of foreground and background components according to their contributions, which results in two feature vectors. The two feature vectors are then fused by deriving weights to generate one feature vector. Furthermore, the feature vector is passed to a CNN classifier for final classification. Experimental results on different classes of normal and abnormal images chosen from different social media outlets and the benchmark dataset show that the proposed approach is effective. In addition, a comparative study with existing methods on the benchmark dataset, a 6-class dataset and another 10-class dataset show that the proposed approach is outstanding in terms of scalability and robustness.

报告人简介:

P. Shivakumara is a Senior Lecturer in Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia. Previously, he was with the Department of Computer Science, School of Computing, National University of Singapore from 2008-2013 as a Research Fellow on video text extraction and recognition project. He has published more than 150 research papers, including TPAMI, PR, CVIU, TCSVT, PRL, IVC, PRL, IJDAR, ICCV, ACMMM, MTA, ESWA, etc.

时间：4月19日 10:00-11:00

地点：计算机科学技术楼225室