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

A multi-label image hashing method with object location awareness

A multi-label and image technology, applied in the field of computer vision, to achieve the effect of eliminating background interference and improving accuracy

Active Publication Date: 2022-04-12
SUN YAT SEN UNIV
View PDF10 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, most of the feature extraction methods ignore the influence of the background on the feature expression, and the retrieved target objects are often included in the complex background in daily pictures, so the feature expression is improved by locating the target object and filtering the background. A study has considerable significance for improving the accuracy of image search

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 multi-label image hashing method with object location awareness
  • A multi-label image hashing method with object location awareness
  • A multi-label image hashing method with object location awareness

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] Such as Figure 1-2 As shown, a multi-label image hashing method with object location awareness includes the following steps:

[0035] S1: collect training sample data;

[0036] S2: Input a picture of 448×448 size into the convolutional subnetwork. The convolutional subnetwork structure here uses the modified GoogLeNet. We remove the last pooling layer in the original structure and add a new convolution kernel size. It is a 3×3 convolutional layer, and the final output is a feature map of 14×14×480;

[0037] S3: A 1×1 convolutional layer is added on top of the feature map obtained in step S2 to obtain a feature map with a size of 14×14, and then softmax operation and truncation operation are performed, and if it is greater than the preset parameter θ, it is taken as 1 Otherwise, it is 0, and finally a 14×14 binary feature map is obtained, which is called a binary mask. The area represented by the value 1 is the area with objects, and the value 0 corresponds to the bac...

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 present invention provides a multi-label image hashing method with object location awareness. The self-learning background filtering structure proposed by the method optimizes the features extracted by the model, and can effectively eliminate background interference, and uses an integrated The trained network structure improves the accuracy of image search.

Description

technical field [0001] The invention relates to the field of computer vision, and more specifically, to a multi-label image hashing method with object position awareness. Background technique [0002] With the rapid growth of the amount of image data on the Internet, how to make full use of the value of these image resources has become an important issue of concern. Research focused on how to query similar pictures in millions or even tens of millions of pictures. The learning-based hashing method is to learn a compressed binary hash code representation of pictures with similar semantics, so that similar pictures also have similar binary hash codes, benefiting from its huge advantages in computing and storage , which has become the mainstream method for large-scale image retrieval. [0003] In recent years, with its powerful learning ability, deep learning has achieved good results in many fields of computer vision, including image recognition, object detection, image segm...

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): G06F16/953G06F16/583G06N3/04
CPCG06N3/04
Inventor 杨尚明潘炎
Owner SUN YAT SEN UNIV