A method and device for image classification based on convolutional neural network

Abstract

The invention discloses an image classification method and device based on a convolutional neural network. The method calculates the neural network weight corresponding to each category of image by receiving input image samples of multiple categories; adopts a layered structure Distributing the corresponding neural network weights of multiple category images, each layer forms a corresponding learning library; processing the input test category image sample data to obtain a corresponding one-dimensional feature description, and corresponding to the test category image sample data The one-dimensional feature description and the neural network weights in the learning library are used for feed-forward learning, so as to determine whether the test category is in the learned category image; the structure of the hierarchical distribution solves the problem of the traditional convolutional neural network. The limitation in the number of classifications prevents the over-learning problem of the convolutional neural network, expands the classification ability of the convolutional neural network itself, and improves the accuracy of the classification, making the image classification algorithm have higher performance in the new environment. robustness.

Description

technical field [0001] The present invention relates to the technical field of image classification methods, in particular to an image classification method and device based on a convolutional neural network. Background technique [0002] Existing image classification methods, commonly used classification methods include supervised learning methods such as neural networks and support vector machines, and unsupervised learning methods such as K-means clustering and nearest neighbor methods. The traditional neural network belongs to the supervised learning method, which obtains the neural network weight characteristic description of the learned object on the basis of learning the existing samples, and distinguishes the categories of the learned library in the external environment according to the learned knowledge. However, due to the limited knowledge of the characteristics of the objects learned, the test objects in the external changing environment may exceed the scope of t...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide an image classification method and device based on a convolutional neural network in view of the above-mentioned defects of the prior art, aiming at solving the problem that the existing neural network image classification method has poor classification effect and is prone to The problem of overlearning

Images