Multi-view People Tracking
This paper presents a hierarchical composition approach for multi-view object tracking. The key idea is to adaptively exploit multiple cues in both 2D and 3D, which are mutually complementary while tracking the humans of interests over time.
Person Search by Joint Inference
This paper presents a novel framework for a multimedia search task: searching a person in a scene using human body appearance. We propose a unified framework which jointly models the commonness of people (for detection) and the uniqueness of a person (for identification).
Matching-based Human Re-identification
This project studies a new surveillance task: re-identifying people at a distance by matching body information. By constructing a compositional template for each query individual, the problem is formulated into a graph matching problem and solved by cluster sampling.
Detection-free Multi-Object Tracking
This project studies a detection-free multi-object tracking algorithm. The proposed framework tracks multiple agents via bi-layer spatiotemporal grouping by exploiting rich appearance and motion information in the observed video sequence.
Complex Background Subtraction
This project studies a simple yet effective background subtraction method that learns and maintains the dynamic texture models within spatio-temporal video patches (i.e. video bricks).
Realtime Object-of-Interest Tracking
This project studies a patch-based object tracking algorithm by learning and maintaining Composite Patch-based Templates (CPT) of the tracking target.