Vehicle image classification method based on extrusion and excitation network

Abstract

The invention discloses a vehicle image classification method based on an extrusion and excitation network, and the method comprises the following steps: S1, carrying out the preprocessing of obtained vehicle images with labels, classifying the vehicle images, and randomly dividing the classified vehicle images into a training group and a test group; s2, constructing an extrusion and excitation-based network, carrying out training by using a training data set, and storing a model; and S3, loading the model, inputting a test group image, obtaining an output result through a weight in the model, and outputting an index of a tensor maximum value, namely a final classification result. According to the vehicle type image classification method, the influence of noise channel can be reduced as much as possible by adopting the design of the extrusion and excitation network, the vehicle type image classification accuracy reaches 95%, and the problems that an existing CNN classification network is not ideal in vehicle type image classification effect and low in accuracy are solved.

Description

technical field [0001] The invention mainly relates to the field of computer image classification, in particular to a vehicle image classification method based on a squeeze and excitation network. Background technique [0002] In the field of computer image classification, deep neural networks are well trained to distinguish images of different categories and exhibit excellent performance. However, in some complex scenes, it is difficult for the computer to distinguish the correct image or the accuracy rate will be greatly reduced. [0003] Chinese patent application CN106203330A discloses a vehicle classification method based on a convolutional neural network. The method includes two stages of training and testing. In the training phase, classify the acquired learning samples; train the sample set of known category samples to obtain excellent network model parameters; extract vehicle model features from the trained model, and use the support vector machine built by liblin...

AI Technical Summary

Problems solved by technology

[0005] Based on the technical problems existing in the background technology, the present invention proposes a vehicle image classification method based on the extrusion and excitation network, which has the characteristics of improving the accuracy of vehicle image classification, and solves the unsatisfactory effect of the existing CNN classification network on vehicle classification , the problem of low accuracy

Images