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

Image generation method based on multiple auxiliary information

A technology of image generation and auxiliary information, applied in image enhancement, image analysis, image data processing, etc., can solve problems such as single information training process, achieve the effect of refining object details and improving quality

Pending Publication Date: 2021-06-29
DALIAN UNIV OF TECH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The current mainstream image generation method only uses a single information to participate in the training process of the model

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 multiple auxiliary information
  • Image generation method based on multiple auxiliary information
  • Image generation method based on multiple auxiliary information

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] The specific embodiments of the present invention will be further described below in conjunction with the drawings and technical solutions.

[0027] An image generation method based on multi-auxiliary information, the steps are as follows:

[0028] Step S1: Extract the COCO data set, extract the scene map of each image according to the annotation information of each image, obtain the training set of the field scene data; extract the image corresponding to the image, obtain the corresponding text Information training set;

[0029] Step S11: First initialize the objects and relationships in the scene map, obtain the object initial matrix and the relationship initial matrix, and then initialize embedding into the GCN network, get the updated object matrix and relational matrix, implement the scene map information Embed, get the scene map vector matrix; the GCN network is stacked by five layers of convolutional blocks, each of which consists of full connectivity, RELU layer, fu...

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 belongs to the field of image generation under a computer vision task, and provides an image generation method based on multi-auxiliary information. According to the method, multiple kinds of auxiliary information are used for guiding a model to complete an image generation task for the first time, the generation task is completed in two stages, in the first stage, input of the model is fusion features of scene graph information and text information, image generation is carried out with the scene graph information as a main part, the text information as an auxiliary part and a GAN network model as a prototype, and a rough image is generated; model input in the second stage is text information and output in the first stage, and the purpose is to enrich image details by using the text information and generate a high-quality image. According to the method, training and evaluation are performed by using a real data set, and meanwhile, performance improvement is evaluated by comparing with a current mainstream image generation model.

Description

Technical field [0001] The present invention belongs to the image generation field under the computer visual task, and involves a method of guiding image generation based on a plurality of auxiliary information. Background technique [0002] In daily production and life, this scene can be seen everywhere: poster designers can't understand customer description, leading to long-term invalid communication between customers and poster designers, low efficiency; witness Describe the appearance of the suspect, the public security organ needs to get the suspect's appearance and break according to the description of witnesses; in the house decoration, according to the description of the owner, if you can quickly see the decoration results, the owner's satisfaction with the landlord It will also greatly improve. For a long time, people have pursued the aesthetic requirements, and the image can be impacted from visual, showing the meaning of the text, the text can be seen from the semantic...

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): G06T5/50G06K9/62G06N3/04G06N3/08
CPCG06T5/50G06N3/08G06T2207/10024G06T2207/20081G06T2207/20084G06T2207/20221G06N3/045G06F18/253
Inventor 牛学硕尹宝才孔雨秋
Owner DALIAN UNIV OF TECH