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

Compressed image recognition method based on deep learning

A technology of compressing images and constructing methods, which is applied in the field of compressed image recognition based on deep learning, can solve the problems of exacerbating compressed image feature drift, loss of details, and limiting recognition performance, so as to improve classification performance, enhance features, and improve recognition accuracy. Effect

Pending Publication Date: 2021-12-10
苏州科亿信息科技有限公司
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, common de-artifacting techniques are not designed to restore imperceptible tiny textures, which will affect the statistical consistency of images and limit the improvement of recognition performance.
In addition, the smoothing operation in de-artifacting will lead to the loss of details in low-compression ratio images, which will exacerbate feature drift in compressed images and degrade recognition performance

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
  • Compressed image recognition method based on deep learning
  • Compressed image recognition method based on deep learning
  • Compressed image recognition method based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0031] This embodiment is combined with a lossless picture data set Div2k containing a large number of natural scenes. Among them, the data is saved in a PNG format, a total of 2000 non-destructive original pictures that do not compress as a raw training image. Technical solutions to the present invention (main processes such as figure 1 The shown is specifically described.

[0032] First, design features enhancement modules, such as figure 2 As shown, including the following steps:

[0033] Step 1: Select the desired pre-training classification model and the corresponding feature location.

[0034] In particular, the pre-training classification model can use common classic architecture, such as VGG16 networks, and SQUEEZENET networks. When using the VGG16 network, the CONV2-2 layer of the network is selected as the characteristic enhanced position; when using the SQUEEZENET network, select the first layer of the network as the position of the feature enhancement.

[0035] The VGG...

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 compressed image recognition method based on deep learning; the method comprises the steps: embedding a trained feature enhancement module into a pre-trained classification model, and obtaining a deep neural network model for compressed image recognition; then, inputting a compressed image needing to be identified into the deep neural network model for classification; and finally, outputting a final classification result. The embedded feature enhancement module does not influence any parameter of the original pre-training model, and the feature enhancement module added in the middle can effectively improve the classification performance of the compressed picture, and does not need to consume a lot of time for retraining.

Description

Technical field [0001] The present invention relates to the field of image compression techniques, and more particularly to a depth learning-based compressed image identification method. Background technique [0002] JPEG (Joint Photographic Experts Group) is widely used in mobile devices, its design is to achieve compromise between bit rate cost savings and visual quality by sacrificing high-frequency components that are not perceived, and do not affect human eye as much as possible In the case of a view, reduce the bandwidth of the image transmission and the volume of the local file store. With the development of artificial intelligence technology, captured images are not only used for human perception, but also for visual analysis such as images, such as image classification, and more. Among them, the image classification is based on the semantic information of the image, which is the basic problem of different categories in the computer vision, also the basis of image detecti...

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): G06K9/62G06N3/04G06N3/08G06T5/10
CPCG06N3/08G06T5/10G06T2207/20052G06T2207/20084G06T2207/20081G06N3/045G06F18/24G06F18/214
Inventor 黄炜
Owner 苏州科亿信息科技有限公司
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