Visible-Infrared Person Re-Identification Based on Dual Attention Mechanism
[Objective]Visible-infrared person re-identification is a very challenging image retrieval problem due to the huge modal difference between visible and infrared images.[Method]In order to further reduce the difference between the two modalities and focus on pedestrian information,a network structure based on dual at-tention mechanism is proposed for visible-infrared person re-identification.On the one hand,through the dual at-tention mechanism to mine personal spatial information of different scales and enhance the channel interaction ability of local features.On the other hand,through learning multi-granular feature information through using global and local branches,different granular information can complement with each other to form a more dis-criminating feature.[Result]Experimental results on two public datasets show that the proposed method has a significant improvement compared with the baseline,and shows ideal performance on both the RegDB dataset and the SYSU-MM01 dataset.[Conclusion]The proposed method can provide an effective reference for solving the problem of modal difference of visible-infrared person re-identification in the future.
visible-infrared person re-identificationattention mechanismmitigate modal differences
国家科技期刊平台
NSTL
万方数据
