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

An Image Generation Method Based on Generative Adversarial Networks

An image generation and network technology, applied in biological neural network models, neural learning methods, instruments, etc., can solve problems such as low image quality and unstable training, and achieve the effect of wide coverage

Active Publication Date: 2021-11-30
珠海亿智电子科技有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Least Squares Generative Adversarial Networks (MaoX, Li Q, Xie H, et al. Least Squares Generative Adversarial Networks [J]. 2016.) is aimed at the two defects of low quality of pictures generated by standard GAN and unstable training. Improve

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
  • An Image Generation Method Based on Generative Adversarial Networks
  • An Image Generation Method Based on Generative Adversarial Networks
  • An Image Generation Method Based on Generative Adversarial Networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0046] The present invention will be further described below in combination with specific embodiments.

[0047] Such as figure 1 As shown, an image generation method based on generative confrontation network combines the module based on paired data (ie: P-module) and the module of unpaired data (ie: U-module) to generate images. U-module uses It is used to pre-train multiple models to solve the problem of label mismatch. P-module is used to train paired data to ensure high fidelity and diversity of images. In the forward propagation stage, P-module can be used The backbone network of [4] generates a large number of license plate images. Such as Figure 4a and Figure 4b As shown, there are various kinds of license plate images generated, which can cover every area, ensure the high fidelity of the image, and combine the module of paired data (ie: P-module) and the module of unpaired data (ie: U-module) The objective function of is: combine the GAN (generating confrontation...

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 present invention proposes a method for generating an image based on a generative confrontation network, comprising the following steps: 1. Input the synthetic data into the generator of the paired data module to train the generated data of the paired data module, and input it into the unpaired data module The generator of the data module is to train the generated data of the unpaired data module; 2, form mixed data; 3, input the mixed data to the discriminator of the paired data module, and use the generated data of the paired data module The data is also input to the discriminator in the paired data module; 4. The judgment result of the discriminator of the paired data module is fed back to the generator of the paired data module. The Inception score of the generated image is close to the real image, and a lower FID score can be obtained, which improves the fidelity and diversity of the data.

Description

technical field [0001] The invention relates to the technical field of image generation of computer vision, in particular to an image generation method based on a generative confrontation network. Background technique [0002] The license plate number is the unique identification of the vehicle, and its particularity and importance determine that the license plate recognition system has become an indispensable and important part of the intelligent traffic management system. The license plate recognition system provides convenient, fast and applicable means for urban traffic management, and has become a research hotspot in recent years. At present, the very popular license plate recognition algorithm is based on deep neural network, so if you want to obtain relatively high accuracy and robustness, you need a lot of license plate labeling data. However, personal privacy information may be involved and there are many types of license plates, which require a lot of time, money ...

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 Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/04G06V20/00G06V20/625G06F18/24G06F18/214
Inventor 殷绪成孙明杨春
Owner 珠海亿智电子科技有限公司