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

A Sketch Reduction Method for Deep Convolutional Neural Networks Based on Perceptual Loss

A neural network and deep convolution technology, applied in the field of sketch simplification of deep convolutional neural networks, can solve the problems of blur, line deformation and distortion, loss, etc., to achieve strong robustness, good visual effect, and wide applicability.

Active Publication Date: 2021-12-21
SOUTH CHINA UNIV OF TECH
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and provide a sketch simplification method based on a deep convolutional neural network based on perceptual loss, which can well simplify sketches of any size and break through the previous sketch simplification Problems such as loss, blur, line deformation and distortion in the method, insufficient simplification ability, etc.

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 Sketch Reduction Method for Deep Convolutional Neural Networks Based on Perceptual Loss
  • A Sketch Reduction Method for Deep Convolutional Neural Networks Based on Perceptual Loss
  • A Sketch Reduction Method for Deep Convolutional Neural Networks Based on Perceptual Loss

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0056] Such as figure 1 As shown, the sketch simplification method of the deep convolutional neural network based on perceptual loss described in this embodiment, the specific method is as follows:

[0057] 1) Obtain the sketch data set, and then obtain its corresponding manual annotation, that is, manually draw a clean line draft on the basis of the sketch, and divide it into a training data set and a verification data set.

[0058] The obtained sketches and corresponding manual annotations are shown in Figure 2(a) and Figure 2(b):

[0059] 2) Convert the image and label data of the image data set into the format required for training the deep convolutional neural network through preprocessing, including the following steps:

[0060] 2.1) Convert the image into a grayscale image;

[0061] 2.2) Flip the images in the data set horizontally, vertically, and diagonally, and do the same flip operation for the corresponding line drawings;

[0062] 2.3) Add tone transformation, ...

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 sketch simplification method of a deep convolutional neural network based on perceptual loss, comprising the steps of: 1) data acquisition; use of draft pictures and corresponding label data; 2) data processing; The draft is converted into the format required for training the deep convolutional neural network through preprocessing; 3) Model construction; according to the training goal, construct a deep convolutional neural network suitable for sketch simplification; 4) Define the loss function; 5) Model training; calculate the loss value of the network according to the loss function, then calculate the gradient of the parameters of each network layer through back propagation, and update the parameters of each layer of the network through the stochastic gradient descent method; 6) Model verification: use the verification data set to verify the training obtained model to test its generalization performance. The method proposed by the invention enables the sketch simplified network to deal with sketches with rougher miscellaneous lines and unclear main structure lines, and has strong robustness to the influence of light.

Description

technical field [0001] The invention relates to the technical field of computer image processing, in particular to a sketch simplification method based on a deep convolutional neural network based on perceptual loss. Background technique [0002] Sketching is the first step in creating art. In this step, creators usually only want to express the concept and composition, rather than focusing on the details. After creating the sketch, the creator needs to draw a clean line draft on the basis of the sketch, which is a tedious and labor-intensive job. [0003] In recent years, with the rapid development of deep learning, deep convolutional neural networks can be used to learn the characteristics of clean line drawings, thereby transforming sketches with many messy lines into simplified line drawings with the characteristics of line drawings. However, the existing problem is that the current deep learning method directly measures the loss by calculating the Euclidean distance b...

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): G06T5/00G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06T5/00
Inventor 徐雪妙谢敏珊缪佩琦
Owner SOUTH CHINA UNIV OF TECH
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