Pixel-level handwriting Chinese-character automatic-generation method, storage equipment and processing device

Abstract

The invention relates to a pixel-level handwriting Chinese-character automatic-generation method, storage equipment and a processing device. The method includes: inputting a preset input vector and an input Chinese character label value to a final generative network model, which is optimized through iterative training, to obtain a handwriting Chinese-character image corresponding to the Chinese character label value. Generation of the final generative network model includes the steps of: acquiring each pseudo-sample on the basis of a pre-built generative adversarial network which includes a generative network model and a discriminative network model and initialization for input vectors and a value of a label corresponding to each sample; and iteratively carrying out parameter optimization of the discriminative network model and correction of the generative network model until iteration conditions are met, and then outputting the final generative network model. The method can directly generate a specified handwriting Chinese-character sample at a pixel level from a random vector, and has the advantages of no need for support of a component library, extensive acquisition and fully automatic generation, diversified styles and the like.

Description

technical field [0001] The invention relates to the technical fields of pattern recognition and machine learning, in particular to a method for automatically generating pixel-level handwritten Chinese characters, a storage device, and a processing device. Background technique [0002] Chinese character sets are large in number, complex in structure, and varied in handwriting styles. A large number of handwritten Chinese character data samples are helpful for the research of handwriting recognition. They can provide a large number of training samples for machine learning and pattern recognition, and can also be used for personalized fonts, However, it is difficult to obtain real handwritten Chinese character data samples, and generally requires a lot of preprocessing operations. [0003] The framework of the traditional simulated handwritten Chinese character generation algorithm is: stroke extraction, stroke matching, statistical structure information, and stroke synthesis. ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problems of cumbersome steps of traditional algorithms an

Images