Fusion Self-attention Feature Clustering Mechanism Network for Person ReID

For the problem that pedestrian features cannot be sufficiently extracted in person re-identification, a person re-identification model based on attention mechanism is proposed. Firstly, pedestrian features are extracted using a hybrid network combining Transformer’s core multi-headed self-attention module with the convolutional neural network ResNet50-IBN-a; Secondly, an self attention mechanism is embedded to make the model of this paper more focused on the key information in the pedestrian foreground; Finally, fusing the mid-level and high-level features in the model can avoid some discriminative features loss. The experimental results show that the provide model achieves 94.8% Rank-1 and 84.5% Rank-1 on the Market1501 dataset and the DukeMTMC-reID dataset, while mAP achieves 84.9% and 65.9%.The model in this paper compares well with some of the existing person re-identification models on all the three main datasets mentioned above.