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

Fine-grained image classification method and device based on image block scoring

A classification method and image block technology, applied in character and pattern recognition, instruments, information technology support systems, etc., can solve the problems of large differences within classes and high similarity between classes, and achieve simple implementation, obvious effects, and improved effects Effect

Active Publication Date: 2022-04-12
ZHEJIANG LAB
View PDF9 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a fine-grained image classification method and device based on image block scoring, which mainly solves the problem of fine-grained image classification with large intra-class differences and high inter-class similarity

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 device based on image block scoring
  • Fine-grained image classification method and device based on image block scoring
  • Fine-grained image classification method and device based on image block scoring

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0038] This method uses the Pytorch framework for experiments, and uses the SGD optimizer with an initial learning rate of 0.03 and a momentum of 0.9 on the CUB bird data. During the training process, the image size is adjusted to 600*600, and it is randomly cropped to 448*448. At the same time, the brightness of the image is randomly fluctuated by 50% on the original basis, the contrast is randomly fluctuated by 50% on the original basis, the saturation is randomly fluctuated by 40% on the original basis, and the image is randomly flipped horizontally. After the pixel value range of the image data is adjusted to 0-1, normalization operations are performed for the R, G, and B channels with mean values ​​of 0.485, 0.456, and 0.406, and variances of 0.229, 0.224, and 0.225, respectively. The training and finetune (fine-tuning) process unify the distributed training of four GPUs, the batch size of each GPU is 8, and the number of training steps is 10,000. During the training proc...

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 device based on image block scoring. The method comprises the following steps: S1, constructing a classification data set; s2, constructing a local identifier; s3, constructing a classification identifier; s4, obtaining image classification feature information and feature information of each image block; step S5, constructing a relevance local identifier; s6, generating a selection identifier; step S7, performing feature processing; step S8, training; and step S9, splicing the global identifier and the associated local identifier, and training the last transformer layer. According to the method, the problem of fine-grained image classification with large intra-class difference and high inter-class similarity degree is solved, plug and play in a transform network is realized, and the effect is obviously improved.

Description

technical field [0001] The present invention relates to the technical field of image classification, in particular to a fine-grained image classification method and device based on image block scoring. Background technique [0002] Fine-grained classification is based on the same large category of images, and carries out finer subcategory divisions, such as the division of bird species, clothing styles, dog species, etc. In real life, there is a huge application demand for identifying different subcategories, such as identifying different types of organisms in the ecological environment for more efficient ecological protection; in the retail industry, automatic identification of the number of times products are taken / tried by customers, statistics Data supports product iteration; in the monitoring industry, better urban management can be achieved by classifying the types of passing vehicles. The wide application value of fine-grained classification tasks has made it a hot r...

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): G06V10/764G06V10/774G06K9/62
CPCY04S10/50G06V10/82G06T7/00G06N3/08G06V10/764G06V10/774G06N3/045
Inventor 苏慧程乐超杨非鲍虎军宋明黎
Owner ZHEJIANG LAB