Abstract
Person Re-identification (Re-ID) is aimed at solving the matching problem of the same pedestrian at a different time and in different places. Due to the cross-device condition, the appearance of different pedestrians may have a high degree of similarity; at this time, using the global features of pedestrians to match often cannot achieve good results. In order to solve these problems, we designed a Spatial Attention Network Guided by Attribute Label (SAN-GAL), which is a dual-trace network containing both attribute classification and Re-ID. Different from the previous approach of simply adding a branch of attribute binary classification network, our SAN-GAL is mainly divided into two connecting steps. First, with attribute labels as guidance, we generate Attribute Attention Heat map (AAH) through Grad-CAM algorithm to accurately locate fine-grained attribute areas of pedestrians. Then, the Attribute Spatial Attention Module (ASAM) is constructed according to the AHH which is taken as the prior knowledge and introduced into the Re-ID network to assist in the discrimination of the Re-ID task. In particular, our SAN-GAL network can integrate the local attribute information and global ID information of pedestrians without introducing additional attribute region annotation, which has good flexibility and adaptability. The test results on Market1501 and DukeMTMC-reID show that our SAN-GAL can achieve good results and can achieve 85.8% Rank-1 accuracy on DukeMTMC-reID dataset, which is obviously competitive compared with most Re-ID algorithms.
Highlights
The biggest feature of smart city is to make full use of the new generation information technology of all walks of life in the city, so as to improve the efficiency of urban management and the quality of citizens’ life
(2) Obtainment of Attribute Spatial Attention. in the person Re-ID network, feature maps of different locations and sizes are selected and combined with the corresponding size of attention heat maps generated by the attribute classification network; Attribute Spatial Attention Module (ASAM) is constructed to assist the discrimination of Re-ID task
In each ASAM, the first half of the channel features of the activation map that keeps the corresponding location of the Re-ID network is kept unchanged in order to maintain a certain amount of global information and avoid information loss that may be caused by the attention mechanism
Summary
The biggest feature of smart city is to make full use of the new generation information technology of all walks of life in the city, so as to improve the efficiency of urban management and the quality of citizens’ life. Scholars generally adopt the method of obtaining global features to solve the Re-ID problem; that is, only the pedestrian ID label is used, and the loss function constraint is adopted to make the network automatically learn the features that are more discriminative for different pedestrian IDs from the entire pedestrian images [12]. In order to solve these problems, we proposed a SAN-GAL network, which combines pedestrian attribute labels and attention mechanism, and can introduce fine-grained attribute features into the Re-ID network for auxiliary discrimination without additional attribute region labeling. In the person Re-ID network, feature maps of different locations and sizes are selected and combined with the corresponding size of attention heat maps generated by the attribute classification network; ASAM is constructed to assist the discrimination of Re-ID task (3) Design Dual-Trace Network. (1) Locate the Attribute Area. in the pedestrian attribute classification network, the attribute labels are used to guide, and the Grad-CAM algorithm [21] is combined to generate AAH (2) Obtainment of Attribute Spatial Attention. in the person Re-ID network, feature maps of different locations and sizes are selected and combined with the corresponding size of attention heat maps generated by the attribute classification network; ASAM is constructed to assist the discrimination of Re-ID task (3) Design Dual-Trace Network. the pedestrian attribute classification network and Re-ID network are trained jointly to achieve the purpose of information interaction and mutual optimization
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
