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

A training method of image codec based on deep neural network

A deep neural network and training method technology, applied in the direction of biological neural network models, neural architectures, instruments, etc., to achieve the effect of avoiding unstable training

Active Publication Date: 2021-10-29
ZHEJIANG UNIV OF TECH
View PDF10 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, there are often more than 2 error functions and more than 3 functional modules in complex image codecs

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
  • A training method of image codec based on deep neural network
  • A training method of image codec based on deep neural network
  • A training method of image codec based on deep neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0020] The present invention will be further described below in conjunction with the accompanying drawings.

[0021] refer to Figure 1 ~ Figure 3 , a kind of training method of the image codec based on deep neural network, described training method comprises the following steps:

[0022] The first step, spatial decoupling: used to decouple the codec and the generation model, and decouple the hidden variable encoding and reconstruction module;

[0023] The second step, time divide and conquer: optimize different loss functions and use different learning rates at different stages of training the codec to improve the speed and stability of training.

[0024] Further, the spatial decoupling aggregates mutually interfering loss functions in the codec into a module, and optimizes the loss function by module during training.

[0025] Furthermore, the modules aggregated according to the spatial decoupling method are decoupled, that is, when a certain module is optimized, it will no...

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

A training method for an image codec based on a deep neural network, the training method comprising the following steps: the first step, spatial decoupling: for decoupling the codec and the generation model, and decoupling hidden variable encoding and reconstruction Coupling of modules; the second step, time division: optimize different loss functions and use different learning rates at different stages of training codecs to improve the speed and stability of training. The invention provides a training method of an image codec based on a deep neural network that can effectively avoid mutual interference of multiple error functions.

Description

technical field [0001] The invention belongs to the field of image codecs, in particular to a training method for image codecs based on a deep neural network. Background technique [0002] For image codecs based on deep neural networks, it is usually necessary to optimize multiple loss functions at the same time during network training, such as reconstruction error functions and image generation against error functions. At the same time, in practical applications, other loss functions will be additionally optimized according to specific needs. These different loss functions have a significant coupling relationship, which will cause serious conflict problems in network training. If the proportion of different error functions is not appropriate, it will lead to unstable training, affecting the reconstruction accuracy of the decoder for the image and the simulation of the generated image, that is, affecting the similarity between the coded image and the training image set . ...

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): G06K9/62G06N3/04
CPCG06N3/045G06F18/214
Inventor 周乾伟陶鹏陈禹行詹琦梁胡海根李小薪陈胜勇
Owner ZHEJIANG UNIV OF TECH