Model training method and device, prediction method and device, re-identification model and electronic equipment

Abstract

The invention provides a model training method and device, a prediction method and device, a re-identification model and electronic equipment. The model training method comprises: obtaining training sample data; inputting the training sample data into the initial model to obtain a category prediction numerical value of the global feature, a category prediction numerical value of the local feature and a global feature vector; calculating a first loss value based on the category prediction value of the global feature, calculating a second loss value based on the category prediction value of the local feature, and calculating a third loss value based on the global feature vector; and optimizing model parameters of the initial model based on the first loss value, the second loss value and the third loss value to obtain a re-identification model. Through the mode, the accuracy of the trained re-identification model on vehicle re-identification can be improved, and the generalization ability of vehicle re-identification is effectively improved through the category prediction numerical value of the global feature, the category prediction numerical value of the local feature and the model training based on the global feature vector.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular, to a model training method, prediction method, device, re-identification model and electronic equipment. Background technique [0002] In recent years, technological developments in the field of computer vision and technological breakthroughs in the field of the Internet of Things have facilitated the realization of the smart city concept. As an important object in smart city applications, vehicles have attracted widespread attention. Researchers have conducted many studies on vehicles. At present, a cutting-edge and important research topic is vehicle re-identification. Vehicle re-identification is one of the core technologies of the intelligent transportation system and is crucial to the construction of a smart city. Through the ubiquitous monitoring network, the vehicle re-identification system can quickly obtain the vehicle that matches the target object. W...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the embodiments of the present application is to provide a model training method, prediction method, device, re-identification model, and electronic equipment to improve "the current vehicle re-identification method lacks the ability to capture finer-grained visual differences, and the algorithm has poor robustness." "The problem

Images