Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Deep learning-based image searching method and device

A deep learning and image technology, applied in the field of image search, can solve the problems that cannot meet the needs of modern Internet applications, low image search accuracy and speed, low search algorithm performance, etc., to achieve fast retrieval speed, high accuracy, The effect of meeting application needs

Inactive Publication Date: 2022-03-11
CHENGDU SEFON SOFTWARE CO LTD
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a method and device for image search based on deep learning to solve the problems of insufficient data and low performance of search algorithms in existing image search methods, which lead to low image search accuracy and speed, and cannot The problem of meeting the needs of modern Internet applications

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
  • Deep learning-based image searching method and device
  • Deep learning-based image searching method and device
  • Deep learning-based image searching method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0058] Such as figure 1 As shown, a deep learning-based image search method is based on the deep learning ResNET50 network, using a large number of images to train the image feature extraction network model, and then loading the model to extract features from the images in the image database to realize image search The function of the same or similar pictures; the specific steps are as follows:

[0059] Step 1, ResNET50 network model construction: use the DL4J deep learning framework to build a ResNET50 network model.

[0060] Step 2, model training: use the ImageNet2012 dataset to train the ResNET50 network model.

[0061] Step 3, image index database design: the fields of the table include "id" is the serial number of the image, "locaPath" is the storage path address of the image on the server, "imgName" is the name of the image and "imgFeature" is the character string of the image feature vector. Table creation statement: CREATE TABLE `image_features` ( `id` int NOT NULL ...

Embodiment 2

[0071] A device for searching images by images based on deep learning includes a memory: used to store executable instructions; and a processor: used to execute the executable instructions stored in the memory to implement a method for searching images by images based on deep learning.

Embodiment 3

[0073] This embodiment is to realize some function codes of this scheme:

[0074] / **

[0075] * Calculate the image feature feature of the input image imgFile

[0076] *

[0077] * @param imgFile

[0078] * @return feature

[0079] *

[0080] * * /

[0081] public String extractImgFeature(File imgFile, ComputationGraphmodelPretrained) throws IOException {

[0082] / / list to store image features

[0083] list imgFeatures = new ArrayList();

[0084] / / Load the ResNet50 weight file trained by ImageNet

[0085] / / ComputationGraph modelPretained = ModelSerializer.restoreComputationGraph(new File(modelPath));

[0086] / / ComputationGraph modelPretrained =loadResNet50PretrainedModel();

[0087] / / System.out.println("model summary:\t" +modelPretained.summary());

[0088] / / load image

[0089] NativeImageLoader nativeImageLoader = new NativeImageLoader(224, 224,3);

[0090] INDArray imgIndarray = nativeImageLoader.asMatrix(imgFile);

[0091] / / model feed f...

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 deep learning-based image searching method and device, and mainly solves the problems of low image searching accuracy and speed and incapability of meeting modern internet application requirements caused by the problems of insufficient data, low searching algorithm performance and the like of an image searching method in the prior art. According to the deep learning-based image searching method, firstly, based on a deep learning ResNET50 network model, all preprocessed images in an image library and to-be-searched images are loaded into the ResNET50 network model to extract feature vectors, and an image feature index database is established through the feature vectors of the image library; then calculating the feature vector similarity between the preprocessed to-be-retrieved picture and all pictures in the picture feature index database, and finally obtaining a set number of high-similarity option return results; the function of searching the same and similar pictures through pictures is achieved, and the method is more intelligent and high in searching accuracy and speed.

Description

technical field [0001] The present invention relates to the technical field of searching images by images, in particular, to a method and device for searching images by images based on deep learning. Background technique [0002] In recent years, with the continuous development of deep learning algorithms and frameworks, artificial intelligence technology has provided new solutions and solutions for the difficulties and pain points existing in traditional industries and fields; especially in the field of computer vision, through the volume of deep convolutional network Multilayer extraction of image features can effectively obtain image feature information on the basis of a large number of training sets, and is more efficient and intelligent than traditional algorithms; however, in the field of image search, due to insufficient data and low performance of search algorithms, leading to The accuracy and speed of image search are low, which cannot meet the needs of modern Inter...

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): G06F16/532G06F16/583G06V10/74G06V10/82G06N3/04G06N3/08
CPCG06F16/532G06F16/583G06N3/084G06N3/045G06F18/22
Inventor 韩威宏刘俊良王怡君张国兵马华均印龙兵刘智勇
Owner CHENGDU SEFON SOFTWARE CO LTD
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
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