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

Image redirection method based on supervised deep network learning

A deep network and redirection technology, applied in the field of image redirection research based on supervised deep network learning, can solve the problem of different execution units.

Pending Publication Date: 2020-11-10
TIANJIN UNIV
View PDF0 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the execution units used for different methods of redirection are not exactly the same, so it is unlikely to combine different redirection methods of execution units, which leads to its result that will not always be able to achieve better performance on a wider range of images. it is good

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 redirection method based on supervised deep network learning
  • Image redirection method based on supervised deep network learning
  • Image redirection method based on supervised deep network learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0050] In order to more clearly illustrate the purpose, technical solutions and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and examples. Apparently, the examples described below are only some embodiments of the present invention, and are not exhaustive of all the embodiments. And, in the case of no conflict, the features in the implementation modes and the examples in this description can be combined with each other.

[0051] The present invention proposes an image redirection method based on supervised deep network learning, the main steps of which include: establishing a new data set, constructing a network model, designing a loss function, model training and model testing.

[0052] Step 1: Create a new data set, which can be divided into five small steps:

[0053] (i) Determine the original input image X. Normalize the collected COCO original image O (3537 images) to a si...

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 relates to an image redirection method based on supervised deep network learning, and the method comprises the steps: selecting and determining an original input image, carrying out redirection operation, evaluating a score for the redirected image, solving an image corresponding to the highest score of each group in a score set, and forming a final new data set; dividing a trainingset and a test set; constructing a generative adversarial network model based on UNet; designing a loss function to measure the difference between the generated redirected image and the correspondingimage in the ground truth set; training the constructed network model by using the newly created data in batches, and continuously optimizing the model through an error back propagation algorithm; andtesting an image in the newly created test set by using the model stored in the training process.

Description

technical field [0001] The technical fields involved in the present invention include computer vision, computer image processing and deep learning, etc., wherein the present invention focuses more on the image redirection research based on supervised deep network learning. Background technique [0002] Image retargeting is designed to resize an image to best fit the target display device. In the past decade, the redirection problem has been extensively studied and many content-aware methods have been proposed. However, these methods can only use low-level semantic features and perform reorientation operations in image space. Therefore, these image retargeting methods often lead to structural distortion and artifacts. [0003] Recently, some deep learning based methods have shown higher performance in image retargeting because they can capture and utilize high-level semantic features in images. However, due to the lack of image retargeting datasets for accurately training ...

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): G06T3/40G06N3/08G06N3/04
CPCG06T3/4046G06N3/08G06N3/045
Inventor 梅怡静潘刚孙迪
Owner TIANJIN UNIV