Training method, system and device for pedestrian re-recognition learning model

Abstract

The invention relates to a training method, system and device for a pedestrian re-recognition learning model. The training method comprises the following steps: collecting all pictures shot by a plurality of cameras from a camera monitoring network; adding an identity label to a picture shot by each camera to form a labeled sample set; utilizing a CycleGAN network to expand the labeled sample set,and carrying out style conversion on pictures shot by each camera, wherein the number of the pictures generated through conversion is N-1, and N is the total number of the cameras, and the expanded pictures are unlabeled samples which form an unlabeled sample set; combining the unlabeled sample set and the labeled sample set to form a training data set; and performing adaptive training through semi-supervised learning according to the training data set by the re-recognition learning model. Compared with the prior art, the training method can effectively reduce the influence of imbalance of the number of positive and negative samples, reduce the cost of needing a large number of label samples in the training process, and enhance the performance of the model.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a training method, system and device for a pedestrian re-identification learning model. Background technique [0002] Pedestrian re-identification, also known as pedestrian re-identification, is a technology that uses computer vision technology to determine whether a specific pedestrian exists in an image or video, that is, given a monitored pedestrian image, retrieve the pedestrian image across devices. Pedestrian re-identification technology can make up for the visual limitations of the current fixed cameras, and can be combined with pedestrian detection and pedestrian tracking technology to be used in video surveillance, intelligent security and other fields. In recent years, video surveillance systems have become widely popular. The construction and application of video surveillance systems are playing an increasingly important role in combating crime and ma...

AI Technical Summary

Problems solved by technology

[0003] At present, some effects have been achieved by using the pedestrian re-identification learning model in the computer to assist manual re-identification, but there are still the following problems: the current pedestrian re-identification learning model training is based on supervised learning, which requires a lot of Labeled data for training, and labeled data is relatively difficult to obtain, it will cost a huge cost of manpower and material resources

In the existing open monitoring network environment, the data of the camera pair is very huge, so obtaining a large amount of labeled data is very costly and almost impossible to achieve, so the correctness and accuracy of the re-identification learning model need to be improved

Images