Unsupervised cross-domain pedestrian re-recognition method based on clustering

Abstract

The invention discloses an unsupervised cross-domain pedestrian re-recognition method based on clustering. The method comprises the steps: firstly inputting a source domain image with a label into a defined network architecture for pre-training, and obtaining a baseline model; then, the label-free target domain image subjected to style conversion being input into a base model for feature extraction and defining a pseudo label, and then the defined pseudo label being used for refining a model pre-trained in the last stage; and finally, loading the trained pedestrian re-recognition model, extracting pedestrian picture features of the to-be-retrieved picture and the target domain, retrieving the most matched pedestrian picture from the target domain, and outputting the most matched pedestrianpicture. According to the method, the practicability of the pedestrian re-recognition model in actual life is effectively improved, the re-recognition performance is improved, the network performanceis good, and the generalization ability is high.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to a cluster-based unsupervised cross-domain pedestrian re-identification method. Background technique [0002] In recent years, person re-identification has been widely studied in the field of computer vision. The goal is to give a picture of a pedestrian to be retrieved, and retrieve the person from the videos taken by several non-overlapping cameras and output it. The existing pedestrian re-identification methods are all based on a priori condition: all pedestrians in the picture have been detected by the detection frame, and the target domain data set is the pedestrian image framed by the detection frame. The original method relies on manually extracting features to label the dataset, which is not only time-consuming and laborious, but also has low performance. In recent years, with the rapid development of deep learning, the performance of pedestrian re-i...

AI Technical Summary

Problems solved by technology

Recently, the method of generating and expanding data sets based on GAN network has been widely studied, mainly to solve the problem of style differences between different cameras, and another application is to expand data sets based on human body pose changes, but human body pose estimation is mostly It is detected based on key points. Due to problems such as alignment, occlusion, and light in the photos taken by the camera, it cannot accurately estimate the pose of the human body.

[0003] Most of the above mentioned are based on supervised learning strategies, which are supervised training on a certain data set. The training effect is better on this data set, but the performance will drop sharply when transferred to other data sets.

In addition, it is time-consuming and laborious to label a large number of data sets in real life, which is basically impossible. Therefore, some researchers recently proposed to use unsupervised domain adaptation methods to improve the pedestrian re-identification model on the unlabeled target training set. performance

Images