Abstract: Particular object retrieval is a fundamental problem in computer vision and machine learning. In this task, discriminative visual representations and efficient matching strategies are two critical requirements. In the first part of my talk, I will review my previous research on generic object retrieval. Based on hand-crafted visual descriptors such as SIFT, two representative works, i.e., the 2D inverted index and query adaptive fusion, will be described. In the second part, I will focus on person retrieval, a specific task in object retrieval. I will share my contribution in conceptually connecting this research field with generic image retrieval, an idea which is widely accepted by the community. Through establishing five large-scale data sets and evaluation protocols, we are directing the community towards learning deep and efficient representations in place of the costly matching practice in the early days. I will also describe our latest contributions in this field based on the generative adversarial networks (GANs), including GAN based network regularization and unsupervised domain adaptation techniques. In the last part, I will brief the research I plan to undertake in the future.
Bio: Dr Liang Zheng is an Assistant Professor in the Singapore University of Technology and Design (SUTD). Prior to this, he was a postdoc researcher in the University of Technology Sydney and University of Texas at San Antonio. He obtained both his B.E. degree (2010) and Ph.D. degree (2015) from Tsinghua University, China. He has published over 20 papers in highly selected venues such as TPAMI, IJCV, CVPR, ECCV, and ICCV. He has made initial attempts in large-scale person re-identification, and his works are positively received by the community. His Google Scholar citations have reached 1600, and the most influential paper is cited 300+ times since 2015. Dr Zheng received the Outstanding PhD Thesis from Chinese Association of Artificial Intelligence, and the Early Career R&D Award from D2D CRC, Australia. His research has been featured by the MIT Technical Review, and 5 papers are selected into the computer science courses in Stanford University and the University of Texas at Austin.