Abstract
Soft biometrics have increasingly attracted research interest and are often considered as major cues for identity, especially in the absence of valid traditional biometrics, as in surveillance. In everyday life, several incidents and forensic scenarios highlight the usefulness and capability of identity information that can be deduced from clothing. Semantic clothing attributes have recently been introduced as a new form of soft biometrics. Although clothing traits can be naturally described and compared by humans for operable and successful use, it is desirable to exploit computer vision to enrich clothing descriptions with more objective and discriminative information. This allows automatic extraction and semantic description and comparison of visually detectable clothing traits in a manner similar to recognition by eyewitness statements. This paper proposes a novel set of soft clothing attributes, described using small groups of high-level semantic labels, and automatically extracted using computer-vision techniques. In this way, we can explore the capability of human attributes vis-a-vis those which are inferred automatically by computer vision. Categorical and comparative soft clothing traits are derived and used for identification/re-identification either to supplement soft body traits or to be used alone. The automatically and manually derived soft clothing biometrics are employed in challenging invariant person retrieval. The experimental results highlight promising potential for use in various applications.