Fast Appearance-based Person Re-identification and Retrieval Using Dissimilarity Representations

TitleFast Appearance-based Person Re-identification and Retrieval Using Dissimilarity Representations
Publication TypeConference Paper
Year of Publication2012
AuthorsSatta, R, Fumera, G, Roli, F
Conference NameConvegno Gruppo Italiano Ricercatori in Pattern Recognition (GIRPR)
Date Published05/2012
Abstract
Person re-identification consists of recognizing an individual who has previously been observed over a camera network. It is a recently introduced computer vision task that can provide useful tools for many applications of video-surveillance. Person re-identification exhibits several challenging issues. Most notable ones are pose variations, partial occlusions, and changing lighting conditions. Another relevant
issue is computational complexity. In this contribution, we present the results attained by the Pattern Recognition and Applications (PRA)
Group in this field: 1) a framework for person re-identification methods; 2) a dissimilarity-based approach for speeding up existing methods; 3) a dissimilarity-based approach for retrieving images of individuals based on a textual query describing clothing appearance, instead of an image.
 
Citation Key 1432
Download: 
AttachmentSize
Satta_GIRPR2012.pdf945.6 KB