Visual Modeling (VISMOD) group focuses on research of innovative machine learning methodologies and technologies to solve real-world computer vision problems including image/video representation, understanding, prediction and enhancement, as well as object (especially human) detection, tracking, recognition and retrieval, oriented to next-generation intelligent video surveillance.
During the past several years, VISMOD group has published many papers in related conferences and journals. Recently, the group members have achieved the 1stplace in ECCV’2018 Pedestrian Detection Challenge, and the Best Paper Award in ICME’2018.
The current main research topics include:
● Models and algorithms in machine learning, especially
1) Metric learning
2) Few-shot learning, meta learning
3) Unsupervised/Semi-supervised/Weakly supervised learning
4) Deep learning
● Object detection, tracking and recognition
1) Person re-identification
2) Object (especially human) detection
3) Single/Multiple object tracking
● Image/video representation learning
1) Attribute learning
2) Visual relationship detection
3) Combinatorial semantic learning
● Image/video generation and enhancement
1) Image generation
2) Video prediction
3) Image/video super-resolution