Image classification system based on cascaded feature blocks

Abstract

The invention discloses an image classification system based on cascaded feature blocks, which comprises a width learning system and a training module for training the width learning system; the width learning system comprises a feature block group, the feature block group is composed of a plurality of feature blocks cascaded in sequence, the output of each feature block is connected with a Flaten layer, the Flatten layer is used for inputting the spliced data into the Top-level Dropout layer, the data processed by the Top-level Dropout layer is input into the Top-level FC layer, and the Top-level FC layer is used for outputting a classification result; the training module is an Adam algorithm. According to the system, the feature blocks are cascaded, so that the depth of the model is increased, the width learning system can learn more abstract information which is more beneficial to discrimination, the feature learning ability of the width learning system can be obviously improved, the classification accuracy is improved, and meanwhile, the training and testing time is still short.

Description

technical field [0001] The invention relates to the technical field of image classification, in particular to a width learning system for image classification. Background technique [0002] Image classification is the process of dividing images into different categories according to their characteristics and specific rules. Image classification is one of the basic and core issues in image understanding. In order to solve the problem of image classification, researchers have proposed many methods, including the current popular deep learning networks including deep belief networks (deep belief networks, DBN), deep Boltzmann machines (deep Boltzmann Machines, DBM) and deep convolution Neural Network (Deep Convolutional Neural Network, DCNN). Because DCNN can learn the advanced features of images, it is widely used in image classification and has achieved good results on complex data sets such as SVHN, CIFAR-10 and CIFAR-100. However, due to the large number of hidden layers o...

AI Technical Summary

Problems solved by technology

[0007] In view of this, the object of the present invention is an image classification system based on cascaded feature blocks to solve the technical problem that the width learning system cannot fully learn the characteristics of the image and the classification performance is poor on more complex data sets

Images