Pedestrian re-identification model training method based on heterogeneous dual networks and feature consistency

Abstract

The invention discloses a domain adaptive pedestrian re-identification model training method, equipment and device based on heterogeneous dual networks and feature consistency, a computer storage medium and a pedestrian re-identification method. A heterogeneous dual-network framework is designed and comprises two asymmetric branches; wherein one network uses convolution with a limited receptive field to obtain local information, the other network uses a Transform module to capture long-range dependence, and mutual learning of heterogeneous dual networks is used to improve heterogeneity and complementarity between the networks, so that robustness to noise pseudo labels is improved; in order to reduce the interference of noise pseudo labels on the network in the optimization process, the feature consistency loss is proposed, the feature consistency loss does not need to depend on any label information, and more attention is paid to the consistency of samples in a feature space; in order to enhance semantic information of the network, a self-adaptive channel mutual sensing module is designed to perform feature extraction on the salient region of the pedestrian, so that the precision and efficiency of pedestrian re-identification are improved.

Description

technical field [0001] The present invention relates to the technical field of machine vision, in particular to a domain-adaptive pedestrian re-identification model training method, equipment, device, computer storage medium and pedestrian re-identification method based on heterogeneous dual networks and feature consistency. Background technique [0002] Person re-identification is a very important research topic in the field of machine vision. Traditional person re-identification mainly uses a large amount of labeled image data for training in specific scenarios. Although supervised learning methods have achieved good results, obtaining labeled data requires a lot of manpower and material resources. In addition, in practical applications, the appearance, background and lighting conditions of pedestrians in different scenes are different, which leads to the model trained on one data set cannot be directly applied to another data set, so how to use It is a research difficult...

AI Technical Summary

Problems solved by technology

[0005] For this reason, the technical problem to be solved by the present invention is to overcome the problem that the training process in the prior art is seriously disturbed by pseudo-label noise

Images