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

A Weakly Supervised Fine-Grained Image Classification Method Based on Hierarchical Feature Transform

A feature transformation, weakly supervised technology, applied in character and pattern recognition, instruments, computing, etc., can solve problems such as large subtle differences, accurate detection and positioning of targets and key area interference, and complex backgrounds of fine-grained images.

Active Publication Date: 2022-07-26
NORTHWESTERN POLYTECHNICAL UNIV
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

At present, weakly supervised fine-grained image classification has achieved excellent performance, but it still faces key difficulties: the complex background of fine-grained images and the subtle differences between subcategories and large intra-class differences give accurate detection and positioning. Severe disruption to targets and critical areas

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 Weakly Supervised Fine-Grained Image Classification Method Based on Hierarchical Feature Transform
  • A Weakly Supervised Fine-Grained Image Classification Method Based on Hierarchical Feature Transform
  • A Weakly Supervised Fine-Grained Image Classification Method Based on Hierarchical Feature Transform

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] The present invention will now be further described in conjunction with the embodiments and accompanying drawings:

[0025] The basic idea of ​​the present invention is: given a fine-grained image data set with only image-level labels, the method uses a training data set, which is combined with the selected candidate regions as a training data set, and trains image-level images on this training data set. Fine-grained image classifier. Then use the image-level feature transformation to co-locate the target in the image, find the area with the most positive correlation in each type of image, and use it as the potential area of ​​the target object, and combine it with the training image as a training data set for training. Object-level fine-grained image classifier. After finding the target latent area, the object-level feature transformation is further used to locate the target component area with the most discriminative ability, and the obtained component area and the t...

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 weakly supervised fine-grained image classification method based on hierarchical feature transformation. By analyzing a data set, only a pre-trained convolutional neural network model is used to co-locate a target in an image. At the same time, the most discriminative components are obtained through negative correlation information analysis. Use images, images and objects, images and parts to train corresponding three-level convolutional neural network classifiers: image-level classifiers, target object-level classifiers, and target component-level classifiers. Granular images for classification. The invention performs feature transformation on the convolution feature, accurately locates the target and the most discriminative component in the fine-grained image, and solves the weakly supervised fine-grained image classification problem with a new idea.

Description

technical field [0001] The invention belongs to the field of computer vision algorithm research, and relates to a weakly supervised fine-grained image classification method based on hierarchical feature transformation, in particular to a method for spatially transforming convolutional features under the category of weakly supervised learning. Methods for fine-grained image classification tasks. Background technique [0002] The problem of fine-grained image classification is a very popular research topic in the field of computer vision in recent years, and its goal is to perform more fine-grained sub-classification of coarse-grained large-class images, such as distinguishing different kinds of birds. Compared with traditional image classification tasks, the differences between fine-grained image classes are more subtle, and can only be distinguished by small local differences. At the same time, illumination, occlusion, posture, background interference, etc. lead to huge intr...

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): G06V10/764G06K9/62
CPCG06F18/241
Inventor 姚西文杨柳青程塨韩军伟郭雷
Owner NORTHWESTERN POLYTECHNICAL UNIV
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