Training method of an object re-identification model and object re-identification method and device

Abstract

The invention discloses a training method of an object re-identification model. The method comprises the steps of obtaining a to-be-trained image set; obtaining an image feature set corresponding to the to-be-trained image set through the to-be-trained object re-identification model; obtaining a category center set; determining a target loss function according to the image feature set and the category center set, the target loss function comprising a first loss function and a second loss function, the first loss function being used for constraining image features corresponding to the same category of images, and the second loss function being used for constraining different category centers; and training the to-be-trained object re-identification model by adopting the target loss functionto obtain an object re-identification model. The invention further discloses an object re-identification method and device. According to the method, the first loss function can be adopted to constrainthe image features of the same category of images, and the second loss function is introduced to constrain different category centers, so that the robustness of the object re-identification model isimproved.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to a training method for an object re-identification model, a method and a device for object re-identification. Background technique [0002] Object weight recognition technology plays an essential role in the fields of intelligent video surveillance, robotics, and autonomous driving. Given an image of an object to be retrieved, object re-identification technology can retrieve related images of the same object from images captured by different cameras. However, the effects of camera perspective, object pose, and occlusion make the object re-identification task quite challenging. [0003] Thanks to the emergence of deep learning technology, object weight recognition technology has developed rapidly in recent years. At present, in the process of training the object re-identification model, the distance constraints between the training samples and the respective maint...

AI Technical Summary

Problems solved by technology

[0004] However, although the currently provided object re-identification model enhances the compactness within the class, so that the features of the same class can be closer to the respective class center features in the Euclidean distance, but ignores the separation between each class center, thus Reduced the robustness of the object re-identification model

Images