PROJECT TITLE :

Hierarchical Graphical Models for Simultaneous Tracking and Recognition in Wide-Area Scenes - 2015

ABSTRACT:

We tend to present a unified framework to trace multiple individuals, likewise localize, and label their activities, in complex long-duration video sequences. To do this, we specialize in 2 aspects: 1) the influence of tracks on the activities performed by the corresponding actors and a couple of) the structural relationships across activities. We tend to propose a two-level hierarchical graphical model, that learns the connection between tracks, relationship between tracks, and their corresponding activity segments, as well as the spatiotemporal relationships across activity segments. Such contextual relationships between tracks and activity segments are exploited at both the amount within the hierarchy for increased robustness. An L1-regularized structure learning approach is proposed for this purpose. While it's well known that availability of the labels and locations of activities can help in determining tracks a lot of accurately and vice-versa, most current approaches have addressed these issues separately. Inspired by analysis in the realm of biological vision, we have a tendency to propose a bidirectional approach that integrates each bottom-up and prime-down processing, i.e., bottom-up recognition of activities using computed tracks and prime-down computation of tracks using the obtained recognition. We demonstrate our results on the recent and publicly offered UCLA and VIRAT data sets consisting of realistic indoor and outdoor surveillance sequences.