EU Science Hub

People Search with Textual Queries about Clothing Appearance Attributes

Abstract: 
Person re-identification consists of searching for an individual of interest in video sequences acquired by a camera network, using an image of that individual as a query. Here we consider a related task, named appearance-based people search, which consists of searching images of individuals using a textual description of clothing appearance as a query, given by a Boolean combination of predefined attributes. People search can be useful in applications like forensic video analysis, where the query can be obtained from a eyewitness report. We propose a general method for implementing people search as an extension of a given reidentification system that uses any multiple part-multiple component appearance descriptor. To this aim, we chose a predefined set of attributes by taking into account the information provided by the original descriptor, transform such a descriptor into a dissimilarity one, and use the resulting dissimilarity values as input features for training attribute detectors.We experimentally evaluate our method on a benchmark re-identification data set.