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

Image Classification Method and System

A classification method and image technology, applied in the direction of instruments, computing, character and pattern recognition, etc., to achieve the effect of improving classification accuracy

Active Publication Date: 2021-12-21
SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, there is currently no method or system for using multi-scale image information and visual saliency to obtain multi-dimensional information of images and perform image classification

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 Classification Method and System
  • Image Classification Method and System
  • Image Classification Method and System

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0037] The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

[0038] refer to figure 1 Shown is a flow chart of a preferred embodiment of the image classification method of the present invention.

[0039] Step S1, using the trained deep convolutional network to automatically extract the high-level semantic features of the original image through layer-by-layer abstraction, and then using the last layer of softmax classifier in the deep convolutional network to predict the The probability value that the original image belongs to each class. In this embodiment, the deep convolutional network that inputs the original image is called the original image network, that is, the first-level network. Wherein, the categories are represented by K={1, 2, . . . , k}. in particular:

[0040] The convolutional network structure adopted in this embodiment (please refer to figure 2 ), including 1 input layer...

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 relates to an image classification method, comprising: predicting the probability value that the original image belongs to each category; judging whether to open a low-resolution network; downsampling the original image to obtain a low-resolution image, and predicting the low-resolution network The probability value that the resolution image belongs to each category; judge whether it is necessary to open the salient area network; combine the prediction results of the original image network and the prediction results of the low-resolution network to obtain the image category; perform saliency detection on the original image to obtain The salient region image is used to predict the probability value of the salient region image belonging to each category; the prediction result of the original image network, the prediction result of the low-resolution network and the prediction result of the salient region network are fused to obtain the image category. The invention also relates to an image classification system. The invention can utilize multi-scale image information and visual salience to obtain image multi-dimensional information, and improve image classification accuracy.

Description

technical field [0001] The invention relates to an image classification method and system. Background technique [0002] Image classification is widely used in many application domains, such as object recognition, image understanding, content-based image retrieval, etc. In recent years, with the breakthrough of deep learning in the field of image processing, the use of deep learning to study image classification has become a research hotspot. [0003] At present, many convolutional neural networks require the input of the convolutional layer to be a fixed-size image segment (such as 224*224), because the feature of the same dimension needs to be input in the fully connected layer, and the dimension of the feature is determined by the size of the input image segment , the size of the convolutional layer and the pooling layer, and the step size. [0004] When people look at an object, they get more detailed image information when viewed at close range, and more outline infor...

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): G06K9/62
CPCG06F18/2415G06F18/256
Inventor 樊春玲张云姜青山
Owner SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI