Unsupervised pedestrian re-identification method based on sample filtering and pseudo label refining

Abstract

The invention provides a method for establishing an unsupervised pedestrian re-recognition model, and the method comprises the steps: carrying out the training of an image sample data set on a source domain, and obtaining a source domain model; creating a collaborative network and a joint network in the target domain, migrating model parameters of the source domain model to the target domain, and initializing the collaborative network and the joint network; respectively inputting the image sample data set into a first temporary average model and a second temporary average model to obtain two groups of image samples, respectively extracting sample features, and obtaining an average sample feature of the sample features of the two groups of image samples; performing clustering processing on the image sample data set to obtain a clustering center and a pseudo tag; image sample data set segmentation processing is carried out on the image sample data set, and the image sample data set is segmented into a trust set and a noise set; and alternately training the joint network and the cooperative network to obtain a trained unsupervised pedestrian re-identification model. The invention further provides an unsupervised pedestrian re-identification method, electronic equipment and a readable storage medium.

Description

technical field [0001] The present disclosure relates to the technical field of pedestrian re-identification, and particularly relates to a method for establishing an unsupervised pedestrian re-identification model, an unsupervised pedestrian re-identification method, electronic equipment, and a storage medium. Background technique [0002] With the development of deep learning, neural network technology has been applied in more and more fields. As an important branch of computer vision, pedestrian re-identification has also received more and more attention. From the general research direction, the research methods of person re-identification can be divided into supervised methods and unsupervised methods. The supervised research method uses marked data to train the network, which can effectively improve the accuracy of recognition. The supervised pedestrian re-identification method can effectively extract features with strong discrimination based on the marked information....

AI Technical Summary

Problems solved by technology

However, this method does not deal with the noise samples participating in the training, and the existence of noise samples will inevitably hinder the improvement of the model recognition effect

In the above method, directly discarding the samples that are far from the sample center will lose the valuable information contained in the samples; if the samples are not processed, more noise will be introduced. How to deal with the noise samples is based on clustering for unsupervised pedestrian re-identification The problem with the method

Images