Pedestrian Re-identification Model based on Joint learning Method

A pedestrian re-identification(Re-ID) model based on joint learning method is proposed. This paper proposed a two-branch model including global feature branch and global feature branch. The global feature branch uses the spatial pyramid pooling model to average the local feature informations of the image and increase the global information of different receptive fields. The local feature branch uses the dynamical distance optimization algorithm without additional information. Experimental results on public datasets show that the proposed model can effectively learn global and local features for Re-ID and outperforms existing basic person re-identification methods.