Model training method, model training device, terminal equipment and storage medium

Abstract

The invention is suitable for the technical field of artificial intelligence, and provides a model training method, a model training device, terminal equipment and a storage medium. The method comprises the following steps: executing two different data augmentation operations on each sample image in a sample set to obtain a first image and a second image; performing pedestrian re-identification on the first image and the second image based on a to-be-trained re-identification model to obtain a re-identification prediction result; calculating a first loss value, a second loss value and a third loss value of the to-be-trained re-identification model based on the re-identification prediction result; and based on the first loss value, the second loss value, the third loss value and a preset target loss function, calculating to obtain a total loss value of the to-be-trained re-identification model, and based on the total loss value, optimizing the to-be-trained re-identification model until the target loss function converges, and obtaining a trained re-identification model. The re-identification model obtained by the method has relatively high robustness and generalization ability. In addition, the invention also relates to a block chain technology.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, and in particular to a model training method, a model training device, terminal equipment, and a computer-readable storage medium. Background technique [0002] Person re-identification (Re-ID) is an image retrieval technology for finding specific pedestrians across devices, which aims to use computer vision technology to retrieve specific pedestrian images under non-overlapping cameras. In the model training task, due to the complexity of the shooting scene, pedestrian images are greatly affected by factors such as shooting angle, illumination, and posture, which makes the image styles acquired by different cameras in different scenes or even in the same scene quite different; in addition, In the same data set, it is inevitable that there will be problems of high similarity between classes and small differences within classes, resulting in low recognition accuracy and p...

AI Technical Summary

Problems solved by technology

In the model training task, due to the complexity of the shooting scene, pedestrian images are greatly affected by factors such as shooting angle, illumination, and posture, which makes the image styles acquired by different cameras in different scenes or even in the same scene quite different; in addition, In the same data set, it is inevitable that there will be problems of high similarity between classes and small differences within classes, resulting in low recognition accuracy and poor generalization ability of the trained model.

[0003] At present, in order to improve the robustness and generalization ability of the model, most person re-identification algorithms are often based on designing a complex network structure to improve the representation ability of the model, but this method improves the representation ability of the model while reducing the algorithm efficiency and increasing the application difficulty

Images