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

Image generation method based on improved graph convolution network

An image generation, convolutional network technology, applied in the field of image processing, can solve problems such as object overlap, object missing, artifacts, etc.

Pending Publication Date: 2021-03-09
JIANGNAN UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But there are artifacts, overlapping objects, missing objects, etc. in the results

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
  • Image generation method based on improved graph convolution network
  • Image generation method based on improved graph convolution network
  • Image generation method based on improved graph convolution network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0068] In order to verify the effectiveness of this method, experiments were carried out on the Visual Genome dataset. This method uses IS (Inception score) and FID (Fréchet Inception Distance) as quantitative evaluation indicators. The IS evaluation index mainly measures the diversity of images generated by the model. , the larger the IS value, the better the diversity of the generated image; the FID evaluation index mainly measures the quality of the image generated by the model, and the smaller the FID value, the better the quality of the generated image. The word vectors of this method are all pre-trained GloVe word vectors, and the vector dimension is selected as d=300. All words not in the word vector dictionary are randomly initialized with 300-dimensional word vectors uniformly distributed between [-1,1].

[0069] Step 1: Build the MDGCN model

[0070] Step 2: Train the MDGCN model

[0071] Set the hyperparameters, input the training set to the MDGCN model, obtain the...

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 an image generation method based on an improved graph convolutional network. The method comprises the following steps: 1, building an input layer, and enabling words in a sentence to be mapped into a low-dimensional and continuous word vector through a pre-trained word vector; 2, establishing a Bi-LSTM layer, and mixing semantic information through the Bi-LSTM layer; 3, constructing implicit layer representation of a target vector: firstly, syntactically mixing information of a target and other words in a sentence through a GCN layer, and then calculating context representation related to the target by using an attention mechanism; 4, constructing an MDGCN layer: constructing a multi-target dependency graph of the sentence according to the dependency syntax tree, and modeling a plurality of targets of the same sentence by using a graph convolution network according to the multi-target dependency graph; 5, establishing an output layer, converting a dimension by using a full connection layer, and converting the dimension into probability representation through a softmax function; and 6, training the model, and taking a cross entropy error function and L2 weight recession as a loss function together.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to an image generation method based on an improved graph convolutional network. Background technique [0002] Computer vision includes many fields such as image generation, semantic segmentation, and target detection. Among them, guiding image generation through natural language description has always been a challenging task in the field of image generation. In recent years, the emergence of deep learning has promoted natural language description-guided image generation. development and has made great progress. [0003] At this stage, Generative Adversarial Network (GAN) has been widely used in the field of image generation. Image generation guided by text description is a hot research field in recent years. Its main task is to generate a picture corresponding to the description content through a text description. The image generation method guided by text description mainly uses t...

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): G06T11/20G06F40/211G06F40/289G06N3/04
CPCG06T11/206G06F40/211G06F40/289G06N3/044G06N3/045Y02D10/00
Inventor 肖志勇张立柴志雷刘登峰吴秦陈璟
Owner JIANGNAN UNIV
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