Person Re-identification by Attribute-Assisted Clothes Appearance

摘要: Person re-identification across nonoverlapping camera views is a challenging computer vision task. Due to the often low video quality and high position, it difficult get clear human faces. Therefore, clothes appearance main cue re-identify person. It represent using low-level features due its nonrigidity, but daily have many characteristics in common. Based on this observation, we study person by embedding middle-level attributes into classifier via latent support vector machine framework. We also collect large-scale dataset, effectiveness of proposed method demonstrated dataset under open-set experimental settings.