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

Fine-grained image classification method and system, terminal equipment and storage medium

A classification method and a fine-grained technology, which is applied in the directions of instruments, calculations, character and pattern recognition, etc., can solve problems such as poor interpretability, difficult network optimization, and occupancy, so as to enhance performance, improve classification efficiency and classification accuracy, and improve The effect of feature diversity

Pending Publication Date: 2022-05-31
XIAN UNIV OF TECH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Difficult to optimize the network due to specially designed attention modules and loss functions
The other is based on high-order information. These methods consider that the first-order information is not enough to model the difference of images, but use high-order information to encode the distinction. The limitation of these methods is that it takes a lot of computing resources, and poor interpretability

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
  • Fine-grained image classification method and system, terminal equipment and storage medium
  • Fine-grained image classification method and system, terminal equipment and storage medium
  • Fine-grained image classification method and system, terminal equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0092] Such as figure 1 As shown, a fine-grained image classification method includes the following steps:

[0093] S1: Acquiring pictures to be classified;

[0094] S2: Input the picture to be classified into the pre-built diversity feature complementary fusion network, and use the diversity feature complementary fusion network to classify the picture;

[0095] Such as figure 2 As shown, the diversity feature complementary fusion network includes three significant and potential feature modules (Significance and Potential Feature Module, SPFM), and the three SPFM are respectively located at the end of residual module 3, residual module 4 and residual module 5 .

[0096] Such as image 3 As shown, SPFM includes a feature block structure, which can be divided along the width of the feature; a 1x1 convolutional layer, which mainly performs a dimensionality reduction process on the feature; and a feature containing the first parameter and the second parameter. The generalize...

Embodiment 2

[0171] A fine-grained classification method based on a complementary fusion network of diverse features, specifically implemented according to the following steps:

[0172] Step 1. First, the classified pictures are passed into the Diversity feature complementary fusion network (Diversity feature complementary fusion network, DFCF). SPFM) to output the salient feature X s , and the saliency suppression of the salient feature as the latent feature X p Continue to pass to the next layer of the network;

[0173] Step 2, the salient feature X extracted in step 1 s They are respectively passed into the 1x1 convolution Conv() module for one-dimensional transformation;

[0174] Step 3, import the features of the transformed dimensions in step 2 into the Feature exchange fusion module (Feature exchange fusion module, FEFM) to perform a feature fusion to enhance feature diversity;

[0175] Step 4. Perform the global average pooling operation on the output diversity features, and tr...

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 fine-grained image classification method and system, terminal equipment and a storage medium. The fine-grained image classification method comprises the following steps: acquiring a to-be-classified image; inputting to-be-classified pictures into a pre-constructed diversity feature complementary fusion network, and classifying the pictures by adopting the diversity feature complementary fusion network; and obtaining a classification result of the picture. According to the method, the pictures are classified by adopting the diversity feature complementary fusion network, the subtle differences of specific areas are automatically and effectively captured by capturing the saliency parts in the pictures, and the pictures are classified based on the subtle differences, so that the classification efficiency and the classification accuracy during picture classification are effectively improved. The method does not need a bounding box, and end-to-end training can be carried out.

Description

technical field [0001] The invention belongs to the technical field of image classification methods, and relates to a fine-grained image classification method and system, a terminal device and a storage medium. Background technique [0002] The purpose of fine-grained is to perform more fine-grained subcategories of images belonging to the same base category. Such as distinguishing wild birds, cars, etc. Since categories have subtle inter-class differences as well as large intra-class differences, it is difficult to capture subtle differences in specific regions for classification. Although deep learning has facilitated research on computer vision tasks, its application to fine-grained classification remains unsatisfactory due to the difficulty in finding informative regions and extracting discriminative features therein. For pose-variety classes like birds, the classification situation is even worse. [0003] Therefore, how to let CNN locate distinguishable parts and lea...

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): G06K9/62G06V10/764G06V10/82G06V10/80G06V10/771
CPCG06F18/213G06F18/2415G06F18/253
Inventor 廖开阳黄港郑元林章明珠王可儿
Owner XIAN UNIV OF TECH