Person re-identification refers to the problem of recognising a person at a different location after one has been captured by a camera at an early location. Solving the re-identification problem has gained increasing attention over the last five years.
The problem has many synonyms, depending on the kind of application considered. Whereas we have a tracking scenario with only one camera, and the target vanishes for a while due to occlusions, the problem is about re-acquisition, e.g. in domotics applications or in health-care scenarios when the primary aim is to keep the identity of a person while one moves about in his/her apartment. For non-cooperative face recognition, solving the re-identification problem is also valuable in a human-robot interaction scenario, where the identity of the interlocutor is maintained and re-used when needed. In a large public space environment such as airport terminals, re-identification is mostly considered as the task of object association in multi-camera networks, where the goal is to keep track of an individual across cameras with non-overlapping fields of views.
Solving the person re-identification problem can benefit from utilising heterogeneous information by learning more effective attributes, exploiting spatio-temporal statistics, estimating how features map across different cameras, taking into account soft-biometric cues (e.g. height, sex), and considering contextual cues (e.g. baggage, other people nearby).
The aim of Re-Id 2012 is to bring together a wide range of researchers in computer vision, pattern recognition and machine learning areas to share innovative ideas and solutions for addressing different declinations of the re-identification problem, proposing novel datasets for benchmarking, and identifying new challenges.
Best Paper Award: A Best Paper Award will be nominated by the Program Committee. The recipient(s) of the award will be given a cash prize (700 euros) sponsored by our industrial sponsors.