Multi-branch and multi-loss person re-identification integrating efficient channel attention
Aiming at overcoming the shortcomings of the existing person feature extraction methods,we propose a multi-branch and multi-loss person re-identification method fused with ECA.Firstly,the lightweight ECA atten-tion module is embedded in the backbone network ResNet50 to enhance salient features and suppress irrelevant features.Secondly,a multi-branch network structure is designed to extract the global and local features of per-son,and different multi-pool feature extraction methods are adopted for different features to enhance the feature extraction ability of the network.Finally,the three types of loss functions are combined to train the model,and BNNeck is used for optimization,so as to improve the robustness of the model.Experiments on Market1501 and DukeMTMC-reID datasets show that the method proposed in this paper has better results and is better than many classic algorithms in recognition accuracy.
person re-identificationECA attention modulemulti-branch featuremulti-loss combination
万方数据
维普
