Date: Wednesday, April 2nd, 2014
Place: Informatics Teaching Lab (ITL) — Top floor seminar room
Speaker: Taiqing Wang
Title: Person Re-identification by Discriminative Video Fragment Ranking
Person re-identification (re-id) aims to associate the observations of the same person captured by different cameras, typically in a CCTV network. It has received increasing attention from both research and industrial communities. Current person re-identification methods typically rely on imagery appearance representations, and ignore space-time information hidden in image sequences. Appearance representation matching is inherently limited for person re-identification in public scenes due to visual ambiguity. This talk will present a novel model to explore identity-discriminative space-time information for person re-id by automatically selecting and ranking the most discriminative video fragments. The comparative evaluation of our model with the state-of-the-art methods will also be presented on two video re-id datasets.
Taiqing Wang is a visiting PhD student from Tsinghua University.