Ancient Yi-character identification method based on convolutional neural network

Abstract

The invention provides an ancient Yi-character identification method based on a convolutional neural network. The method comprises the steps of constructing the convolutional neural network which is composed of four convolutional layers, two full-connecting layers and one softmax layer, and adding an additional convolutional layer in front of each convolutional layer; combining a ResBlock, an Inception and an SEBlock for forming a mixed structure, and adding the SEBlock behind the Inception structure for replacing a weight layer in an original ResBlock, and adding the mixed structure in frontof each convolutional layer for obtaining an improved convolutional neural network; respectively performing convolution through three convolutional cores of 1*1, 3*3 and 5*5 in the Inception structureand performing channel stacking on three outputs, restoring the channel number by means of a 1*1 convolutional core, performing characteristic re-calibration on the output of the Inception structureby means of the SEBlock, and adding residual errors after re-calibration to the original input. The ancient Yi-character identification method has advantages of reasonable conception, high identification effect and effective prevention of a performance reduction problem in identification.

Description

technical field [0001] The invention relates to the technical field of character recognition, in particular to a method for recognizing ancient Yi characters based on a convolutional neural network. Background technique [0002] Text recognition technology is an important topic in the field of computer vision, which involves many fields such as machine learning, natural language processing, statistics, etc., and has always been a relatively hot topic. The purpose of text recognition is to convert the text in the image into digital form through a series of processing. Character recognition is not a new problem. People tried to study character recognition long before the invention of computers. Traditional text recognition mainly relies on prior rules and artificial features. People often need to perform cumbersome preprocessing and feature extraction on the original image, but these often cannot fully represent a thing, and more or less will miss some more important things. ...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems mentioned in the above-mentioned background technology, the present invention proposes a method for recognizing ancient Yi characters based on a convolutional neural network, which can avoid the problem of performance degradation as the number of convolutional layers increases in the recognition of ancient Yi characters. At the same time, the performance of the model is further improved with a limited amount of calculation; at the same time, a set of sample increment process is proposed to increment the handwritten samples, expand the sample set, improve the stability of the model, improve the performance of the model, and avoid tediousness The preprocessing technology is an end-to-end ancient Yi language recognition method

Images