Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Style-controllable image text real-time translation and conversion method

A conversion method and text technology, applied in natural language translation, instruments, biological neural network models, etc., can solve problems such as model collapse, slow convergence speed, unstable training process, etc., and achieve the effect of improving recognition

Pending Publication Date: 2020-09-29
CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF2 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Traditional style transfer GAN algorithms, such as DualGAN, CyCleGAN, and Pix2Pix, have problems such as slow convergence speed, unstable training process, and prone to model collapse.

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
  • Style-controllable image text real-time translation and conversion method
  • Style-controllable image text real-time translation and conversion method
  • Style-controllable image text real-time translation and conversion method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0050] Below in conjunction with accompanying drawing and specific embodiment the present invention is described in further detail:

[0051] A real-time translation and conversion method of image text with controllable style, such as figure 1 As shown, it is a flow chart of the style controllable image text real-time translation and conversion method of the present invention, the method includes:

[0052] S1, scene image input, input a real scene image containing text, the picture may contain complex background, fence with text features or grid wall tiles, the text may have artistic style, even the text is distorted or have a perspective relationship.

[0053] S2, text position detection, the scene image obtained in the above steps is used as the input of the CRAFT text position detection network, and the text position detection is performed on the scene image, and the text position detection network structure is as follows figure 2 shown. For the input original image, it ...

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 style-controllable image text real-time translation and conversion method. The method comprises the following steps: taking a scene image as input; performing feature extraction by using a multi-layer CNN network, and detecting the position and form information of the image text; and then erasing text pixels based on a text positioning box to obtain a background image anda mask, and carrying out background image restoration by using a thick restoration network and a thin restoration network based on a codec structure; performing form correction and style removal on the image text to obtain a common font image text; and recognizing the image text by using a CRNN model, performing correcting by combining text semantics, and performing translating or converting according to requirements; performing stylization processing on the translated text by learning the artistic style of the original text; and outputting a scene image with a controllable text conversion style. According to the method, more valuable information can be analyzed from the scene image, and the information storage degree during image text translation and conversion is remarkably enhanced.

Description

technical field [0001] The invention belongs to the field of computer graphics and image processing, and relates to a style-controllable image text real-time translation and conversion method. Background technique [0002] With the continuous development of economic globalization and network technology, international exchanges are increasing day by day. The language barrier between people who use different languages ​​has caused cultural barriers, which has a negative impact on international friendly exchanges and the spread of fine culture. Breaking down language barriers and eliminating cultural barriers is an important way to enhance international exchanges. Text symbols are the manifestation of language, and vision is an important source of text information acquisition. Therefore, the real-time recognition, translation and conversion technology for scene text is expected to be high. In recent years, due to the rapid increase in the number of images and the rapid develop...

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
IPC IPC(8): G06F40/58G06F40/30G06K9/46G06K9/20G06N3/04
CPCG06F40/58G06F40/30G06V10/22G06V10/40G06V30/10G06N3/045G06N3/044
Inventor 李昕刘航源王志宽吕涵
Owner CHINA UNIV OF PETROLEUM (EAST CHINA)
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com