Data augmentation method and image classification method based on selection and generation

An image and data technology, applied in the field of image analysis and recognition, can solve problems such as high cost, achieve the effect of expanding quantity and diversity, benefiting detailed features, and good classification and recognition results

Active Publication Date: 2019-04-12
PEKING UNIV
View PDF5 Cites 22 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, labeling requires a lot of manpower

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
  • Data augmentation method and image classification method based on selection and generation
  • Data augmentation method and image classification method based on selection and generation
  • Data augmentation method and image classification method based on selection and generation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0033] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0034] A data augmentation method based on selection and generation applied to image classification in this embodiment, its flow is as follows figure 1 As shown, it specifically includes the following steps:

[0035] (1) Data segmentation

[0036] Using the Selective Search algorithm (Selective Search) to generate thousands of image blocks for each original training image, these image blocks have a certain probability to contain the target object area, and each image block has a corresponding probability score. Then, by pseudo-random sorting (Pseudo Random Sorting), the first N image patches are selected as the expansion of the training image. figure 2 is a schematic diagram of an image block obtained after data segmentation of a picture containing a yellow warbler.

[0037] (2) Data filtering

[0038] Most of the N image blocks ...

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 provides a data augmentation method based on selection and generation and an image classification method, and the method comprises the following steps: carrying out the segmentation of an input image, and generating a plurality of image blocks to increase the number of training images; filtering the obtained image blocks, namely classifying the image blocks by utilizing a convolutional neural network, and selecting the image blocks related to the target object; image blocks obtained through filtering in the last step are reselected through multi-example learning, and image blockscontaining most areas of the object are selected; and finally, learning a corresponding relationship between the image and the text by utilizing the generative adversarial network, generating more new images by utilizing text description, and further expanding the diversity of the training image. According to the method, only one training sample and text description information thereof are used,and the image data diversity is amplified by segmenting, filtering, reselecting and generating the data. Image classification model training is carried out by using the amplified image data, thereby realizing image classification under a training sample condition.

Description

technical field [0001] The invention relates to the technical field of image analysis and recognition, in particular to a data augmentation method based on selection and generation and an image classification method using the method. Background technique [0002] In recent years, with the rapid development of Internet technology and multimedia technology, Internet images have shown explosive growth. Image classification is a difficult research problem in the field of computer vision by analyzing the content of the picture and giving its category information. [0003] Traditional image classification methods mainly include two stages: feature extraction and classifier prediction. In the feature extraction stage, feature extraction is performed on the input image. There are usually two ways of feature extraction: one is intensive feature extraction, and the other is feature extraction for points of interest, such as extracting SIFT key point information , which is further qu...

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
IPC IPC(8): G06K9/62G06N3/04
CPCG06N3/044G06F18/2411
Inventor 彭宇新何相腾
Owner PEKING UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap