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

A Method for Evaluation of Salient Datasets

An evaluation method and data set technology, applied in the field of image processing, can solve problems such as increasing the difficulty of salient object detection

Active Publication Date: 2019-10-29
BEIJING UNION UNIVERSITY
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

When the difference between the foreground object and the whole image is small, it naturally increases the difficulty of salient object detection

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 Method for Evaluation of Salient Datasets
  • A Method for Evaluation of Salient Datasets
  • A Method for Evaluation of Salient Datasets

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0060] Such as figure 1 As shown, step 100 is executed to open the data set to be measured.

[0061] Step 110 is executed to calculate the proportion of the size of the significant region in the entire image.

[0062] An image I and the binary labeled map G corresponding to this image, assuming that the number of salient regions that are not connected to each other in the binary labeled map G is m. x i， 1≤i≤m, representing the salient area of ​​block i in image I, calculate x i The percentage of the area of ​​the entire image. Divide percentages into 10 scale classes, [0,0.1),[0.1,0.2),[0.2,0.3),[0.3,0.4),[0.4,0.5),[0.5,0.6),[0.6,0.7) ,[0.7,0.8]),[0.8,0.9),[0.9,1],x i In which scale level, add 1 to the number of significant regions within the scale range, num j =num j +1,1≤j≤10. Perform the above operation on each image in the data set, and finally calculate the percentage of the number of salient regions in the 10 scale levels to the number of all salient regions.

...

Embodiment 2

[0070] The calculation process of the percentage of the salient area size of the dataset to the entire image is as follows:

[0071] Input: data set D and its corresponding binary label map S;

[0072] Output: The percentage of the number of significant regions within the 10 scale levels to the number of all significant regions.

[0073] calculation process:

[0074] 1.for image i∈[1,10] do

[0075] 2.num i = 0; / / initialize the number of each level to 0

[0076] 3. end for

[0077] 4.for image I j ∈ D do

[0078] 5. Read I j and I j The corresponding binary label graph G;

[0079] 6. Extract the connected salient region set C in the binary labeled graph G, and the number of C is m;

[0080] 7. for x i ∈C, 1≤i≤m do

[0081] 8. Calculate x i image of I j percentage of area

[0082] 9. Judgment i The rank is j, num j =num j +1, 1≤j≤10;

[0083] end for

[0084] 10. end for

[0085] 11.for j∈[1,10] do / / Calculate the ratio of the number of significant region...

Embodiment 3

[0090] The calculation process of the number of salient regions connected to the edge of the image and the proportion of all salient regions in the dataset is as follows:

[0091] Input: data set D and its corresponding binary label map S;

[0092] Output: The number of salient regions connected to the edges of the image and the proportion of all salient regions in the dataset.

[0093] calculation process:

[0094] 1.num represents the number of salient regions connected to the edge of the image, num=0;

[0095] 2.sum represents the number of significant regions, sum=0;

[0096] 3.for image I j ∈ D do

[0097] 4. Read I j and I j The corresponding binary label graph G;

[0098] 5. Extract the connected salient region set C in the binary annotation map G, the number of C is m;

[0099] 6. sum=sum+m;

[0100] 7. for x i ∈C, 1≤i≤m do

[0101] 8. if x i connected to the edge of the image

[0102] 9. num=num+1;

[0103] 10. end if

[0104] 11. end for

[0105] 12. en...

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 provides a method of evaluating the performance of a saliency data set. The method comprises the following steps: the ratio of a salient region to the whole image is calculated; the ratio of the number of salient regions connected with an image edge to all salient regions in the saliency data set is counted; RGB color feature differences between a salient region and the whole image are counted; and the performance score of each saliency data set from the first step to the third step is calculated. The data set can be counted from different angles to comprehensively evaluate the performance of the data set. Research and development on an objective and scientific saliency detection algorithm are facilitated, and design of an algorithm with not high robustness to cater to database deviations can be avoided.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a performance evaluation method of a saliency data set. Background technique [0002] From the relevant literature on saliency datasets, saliency datasets come from two fields: one is a dataset specially established for saliency research, and the other is a saliency dataset extended from the field of image segmentation. At present, many datasets have simple image structures, and the foreground and background have obvious differences, such as color differences, which will lead to easier detection of salient objects in the image. Additionally, many saliency datasets come with significant center bias. However, there are not many evaluation methods for the dataset. [0003] In the paper (Visual Saliency Based on Scale-Space Analysis in the FrequencyDomain.pami.2012), a data set containing 235 images was constructed, and the size of the salient objects in the data set was c...

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): G06T7/11G06T7/90
Inventor 梁晔李华丽陈强宋恒达胡路明蒋元昝艺璇
Owner BEIJING UNION UNIVERSITY
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