Structure search method of image super-resolution generative network

A technology of network structure and search method, applied in biological neural network model, image data processing, graphics and image conversion, etc., can solve the problems of mode collapse and instability of confrontation network training, achieve good performance, enhance training stability, strong The effect of data adaptation performance

Active Publication Date: 2020-03-27
SOUTHEAST UNIV
View PDF4 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is to provide a structure search method of image super-resolution generation network, which can obtain a relatively stable training loss, and improve the problem of unstable training of generative confrontation network and easy mode 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
  • Structure search method of image super-resolution generative network
  • Structure search method of image super-resolution generative network
  • Structure search method of image super-resolution generative network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0049] Such as figure 1 As shown, an image super-resolution generation network structure search method, including the following steps:

[0050] Step S1: Set up the structural search space of the shared generator in the image super-resolution generative network The search space is divided into two categories, which are the search space of the residual convolution unit in the generator and upsampled convolutional unit search space The generator network structure controller in the search space The Network Structure of Medium Sampling Generator

[0051] Step S2: Use the loss stabilizer to train the shared generator sampled by the network structure controller on the small-scale super-resolution image dataset, and obtain the relationship between the high-resolution generated image of the current sampled generator and the high-resolution real image Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM);

[0052] Step S3: Send the peak signal-to-noise rati...

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 structure searching method of an image super-resolution generative network. A network structure controller automatically samples a network structure of an optimal generator without help of experts. The sampled image super-resolution generation network can provide reward signals for the network structure controller according to the performance of the network structure controller, so that the network structure controller continuously updates the parameters of the network structure controller, and a high-resolution generation image can obtain the optimal peak signal-to-noise ratio and structural similarity; and when the size and the characteristics of the image needing to be subjected to the super-resolution operation are changed, searching again through the networkstructure controller to obtain the optimal image super-resolution generation network under the current characteristic form image.

Description

technical field [0001] The invention relates to the technical field of image super-resolution, in particular to a structure search method for an image super-resolution generation network. Background technique [0002] Image super-resolution is a basic and important vision problem, aiming at recovering high-resolution images from low-resolution images. Image super-resolution technology has a wide range of practical application prospects. For example, the details of photos taken by mobile phones are seriously lost in dark light environments, and the lost details can be restored through image super-resolution technology; various websites can compress image and video quality to reduce transmission. When it reaches the client, it will reply high-definition images and videos through super-resolution technology. Existing image super-resolution methods mainly include methods based on traditional interpolation theory, methods based on convolutional neural networks, and methods based...

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): G06T3/40G06N3/04G06N3/08
CPCG06T3/4053G06N3/08G06N3/045
Inventor 莫凌飞管旭辰
Owner SOUTHEAST UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap