Learning Semantic Representation on Visual Attribute Graph for Person Re-Identification and Beyond

Person re-identification (re-ID) aims to match pedestrian pairs captured from different cameras. Recently, various attribute based models have been proposed combine the as an auxiliary semantic information in order learn a more discriminative representation. However, these methods usually directly concatenate visual branch and embeddings final representation, which ignores relation between revealed by similarity. To capture explore such relation, we propose unified representation framework, called Visual Attribute Graph Embedding Network (VAGEN), simultaneously We unify embedding similarity into (VAG), where is considered node edge. Then, graph generate through Neural Network. Except for this embeddings, VAGEN also conducts implicitly hard example mining similar false positive results, has not explored yet among existing methods. conduct extensive empirical studies on several person re-ID datasets evaluate our algorithm aspects. The results show that method outperforms state-of-the-art techniques with considerable margins.