【澳门太阳2007网址】Adaptive Deep Learning and Person Re-identification Method
Topic: Adaptive Deep Learning and Person Re-identification Method
Lecturer: Professor and Ph.D. Supervisor Lai Jianhuang, School of Data and Computer Science, Sun Yat-sen University
Host: Professor Xiao Fen, XTU College of Information Engineering, Intelligent Computing and Information Processing-Ministry of Education Key Laboratory executive deputy director, and China Computer Federation Artificial Intelligence & Pattern Recognition Committee member
Time: 11:10-11:50, November 17 (Saturday), 2018
Venue: Auditorium A 217, XTU College of Civil Engineering and Mechanics
Professor Lai now serves as China Society of Image and Graphics deputy director, China Computer Federation (CCF) Computer Vision Committee deputy director, CCF Artificial Intelligence & Pattern Recognition Committee member, and Chinese Association for Artificial Intelligence Machine Learning Special Committee member. He is also Guangdong Provincial Key Laboratory of Information-Security Technology director, Key Laboratory of Video and Image Intelligent Analysis and Application Technology of Ministry of Public Security deputy director, Guangdong Image and Graphics Society director (4th and 5th sessions), and CCF excellent member and director.
The bottleneck of semi-supervised learning is that there are over-fitting problems when learning with a small amount of tag data, especially in complex deep convolutional neural networks. In order to solve this problem, the semi-supervised learning is mathematically modeled as an EM algorithm.
The Deep Growing Learning (DGL) is used to ensure that when the tagged data is small, the network is in a shallow state; while the number of pseudo tags increases, the network deepens, and thus the over-fitting problem is alleviated. In addition, the report will also introduce the team’s research progress in the person re-identification based on space-time clues, especially the identification components analysis in cross-view, and that based on deep learning.