Associate Professor with Tenure
Research Interests: Multimedia information retrieval
Zhang, Shiliang is an assistant professor in the Department of Computer Science and technology, School of EECS. He obtained his Ph.D. degree from Institute of Computing Technology, Chinese Academy of Sciences in 2012. After that, he was a Postdoctoral Fellow in University of Texas at San Antonio and a Postdoctoral Scientist in NEC Labs America, Cupertino, CA. His research interests include large-scale image retrieval, computer vision for autonomous driving, and deep learning for image understanding.
Dr. Zhang has published more than 40 research papers, and most of them are published in top-tier conferences and journals, such as ACM Multimedia, ICCV, ECCV, IEEE T-PAMI, IEEE T-IP, IEEE T-MM, and Pattern Recognition. He received the Outstanding Doctoral Dissertation Awards from the Chinese Academy of Sciences and Chinese Computer Federation (CCF), the President Scholarship by Chinese Academy of Sciences, NEC Laboratories America Spot Recognition Award, and the Microsoft Research Fellowship. He is a recipient of Top 10% Paper Award in the IEEE MMSP 2011. He served as Technical Program Committee Chair of 2017 1st CVPR Workshop on Target Re-Identification and Multi-Target Multi-Camera Tracking. His research is supported by Natural Science Foundation of China.
Dr. Zhang’s research achievements are summarized as follows:
1) Compact and discriminative visual feature extraction: Lots of research efforts have been made in extracting local descriptors in the past decade. However, those descriptors like SIFT are not compact and efficient enough, making them not applicable to scenarios like mobile computing and mobile visual search. To conquer those issues, Dr. Zhang proposed several compact visual descriptors like Discriminative Visual Words and Visual Phrases, Edge-SIFT, and Ultra Short Binary descriptors that are proven efficient and discriminative in large-scale image applications.
2) Visual content annotation and understanding: Understanding semantics from multimedia data is a key step in multimedia information retrieval. Dr. Zhang has conducted several works that annotate semantic tags for general images, predict semantic attributes for pedestrian images, and understand affective cues from music videos, respectively. Dr. Zhang published one of the earliest works on music video affective analysis and retrieval. His works on semantic annotation show novelties in the aspects of requiring less training data, featuring low computational complexity, and getting competitive performance in public datasets.
3) Large-scale image indexing: Offline indexing not only decides the retrieval efficiency but also affects the accuracy. Dr. Zhang has made many research efforts to fuse multiple features during off-line indexing and to eliminate the data redundancy. These works results in better retrieval accuracy, efficiency, and memory consumption. For example, Dr. Zhang designed a co-indexing strategy that seamless fuses multiple features during indexing stage. This co-indexing strategy does not change the online retrieval procedure but still substantially improves the retrieval efficiency and accuracy.