Unlock instant, AI-driven research and patent intelligence for your innovation.

Text recognition method based on global and local attention mechanisms

A text recognition and attention technology, applied in character and pattern recognition, neural learning methods, computer components, etc., to achieve the effect of improving forward speed, strong robustness, and improving representation ability

Active Publication Date: 2021-11-26
SOUTH CHINA UNIV OF TECH +1
View PDF8 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But these methods have weaknesses in modeling global relations and model forward velocity

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Text recognition method based on global and local attention mechanisms
  • Text recognition method based on global and local attention mechanisms
  • Text recognition method based on global and local attention mechanisms

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0052] refer to figure 1 As shown, the present invention proposes a text recognition method based on a global and local attention mechanism, which specifically includes the following steps:

[0053] S1. Data acquisition: use synthetic text lines and public real text lines as training sample data, and use text lines collected in real scenes as test sample data.

[0054] Among them, the synthetic data is synthesized while training.

[0055] S2. Data processing: perform data preprocessing operations on training samples and test samples, refer to figure 2 shown.

[0056] Wherein, all comprise offline text and online text in the training sample and test sample; For offline text, while keeping the aspect ratio, the offline text picture height is normalized to 128 pixels, denoted as picture I;

[0057] For online text, rotate the online handwriting to be horizontal, and then perform the feature extraction operation, specifically:

[0058] The online handwriting S is a sequence c...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a text recognition method based on global and local attention mechanisms. The method comprises the following concrete steps: taking synthetic data and a public real text line as training data, and taking a text line collected in a real scene as test data; performing data preprocessing on all the training and testing samples; training a recognition model by adopting a supervised method, wherein each text line has corresponding text information; designing a recognition network by adopting structures such as a convolutional layer and a self-attention layer; inputting the training data and the label into a recognition network for training; and inputting test data into the trained network to obtain an identification result. According to the method, text line recognition is realized by utilizing global and local attention networks, complex and diversified online and offline texts in various scenes can be recognized, relatively high accuracy is achieved on each public data set, and the method has extremely high practicability and application value.

Description

technical field [0001] The invention relates to the technical fields of pattern recognition and artificial intelligence, in particular to a text recognition method based on a global and local attention mechanism. Background technique [0002] With the rapid development of computer technology, artificial intelligence technology has also become an indispensable driving force in production and life. In recent years, the advancement of deep learning technology and the development of related hardware have made the practical application of deep neural networks possible. [0003] Text can be seen everywhere in daily life and is the carrier of information transmission. Most of the information that humans get from vision comes from text. At the same time, text is also the carrier of human civilization and knowledge inheritance. Therefore, how to automatically recognize text information from text pictures or online handwriting has become a research of great value and significance. ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06K9/46G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/2415G06F18/214Y02T10/40
Inventor 彭德智金连文谢灿宇李鸿亮
Owner SOUTH CHINA UNIV OF TECH