Title: Image Saliency Detection: From Convolutional Neural Network to Capsule Network

Report time: 9: 30-11: 00, December 26, 2019

Reporting location: Yifu Building 202

Report Content: Human beings possess the innate ability of identifying the most attractive regions or objects in an image. Salient object detection aims to imitate this ability by automatically identifying and segmenting the most attractive objects in an image. In this talk, I will share with you two recent works that we published in the top venues. In the first work, we showcase a guidance strategy for multi-level contextual information integration under the CNNs framework, while in the second work, we demonstrate how we carry out the saliency detection task using new Capsule Networks.

Speaker profile: Junjun Han currently works at the University of Warwick (ranked 62 in the QS world), is an associate professor of data science (a tenure), and leads research in computer vision at the college (supervises 8 full-time doctoral students) . Prior to joining the University of Warwick, he was an associate professor of tenure at Lancaster University (QS 130 in the world) and a tenure at the School of Computer and Communication. Prior to working in the United Kingdom, Dr. Han was a senior scientist at Philips (Civolution) in the Netherlands (2012-2015) and led the company ’s product development efforts as the chief expert in the company ’s AV / audio fingerprint recognition direction. From 2010 to 2012, Dr. Han worked at the Institute of Mathematics and Computers of the Royal Academy of Sciences of the Netherlands and participated in one of the seventh framework research project of the European Union as co-leader of the project. From 2005 to 2010, Dr. Han worked in the Signal Processing Group of the Technical University of Eindhoven in the Netherlands. He has participated in the research of 2 EU projects and directed nearly 10 PhDs and masters.

Han Jungong and his research team published more than 100 SCI journal papers in multimedia content recognition, computer vision, machine learning, etc., with a total impact factor of more than 350. The research results have been published more than 40 times in top journals in this field (such as: IJCV) and mainstream IEEE journals (including 17 IEEE T-IP). In addition, Dr. Han has published more than 40 papers at computer CCF Class A conferences (eg, NeurIPS, ICML, CVPR, ICCV, ACM MM, AAAI, etc.). The total google citation rate of the article exceeded 4,500 times, and the highest citation rate for a single article was 1470 times (IEEE TCYB 2013-2019 citation rate, both the first download rate). 10 papers were included as highly cited papers by Web of Science, and 1 paper was cited. Selected as a hot paper. Associate editor of Elsevier Neurocomputing (IF 4.1) magazine, Springer Multimedia Tools and Applications (IF 2.1) magazine and IET Computer Vision (IF 1.6) magazine and 4 well-known magazines in the field (IEEE Transactions on Neural Network and Learning System, IEEE Transactions on Cybernetics, etc.); he is also a member of the Standing Committee of IEEE Industry DSP Technology and a member of the Technical Committee of IEEE Multimedia Communications.