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

Image set generating method and device and image recognition model training method and system

A technology of image recognition and image collection, applied in the field of image recognition, can solve the problems of limited improvement of model training effect and no significant increase in the dimension of image collection, etc., achieve good data quality, improve generation efficiency, and reduce the influence of background noise and clutter Effect

Inactive Publication Date: 2018-02-13
ALIBABA GRP HLDG LTD
View PDF5 Cites 29 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For example, the image of a cat, after the above transformation, has no increase in the number of features. From the point of view of the feature space, it is still the same image before and after the transformation, and the dimension of the image set has not been significantly improved. The improvement of the training effect of the model is limited.

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
  • Image set generating method and device and image recognition model training method and system
  • Image set generating method and device and image recognition model training method and system
  • Image set generating method and device and image recognition model training method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] This embodiment provides a method for generating an image set for image recognition. The images herein include pictures, and may also include animations, video segments, and the like. Such as figure 1 As shown, the method of this embodiment includes:

[0037] Step 110, obtaining 3D models of multiple categories;

[0038] The categories here can be various according to business needs.

[0039] In the business field involving commodity trading, categories can be divided according to individual items such as beds, sofas, TV sets, shoes, etc., and each item can also be subdivided, such as Nike, Adidas and other brand shoes as a category. At this point the 3D model is a 3D model of these items. A category such as sofas has some common features, and there are also differences within the category (such as leather sofas and cloth sofas), and one or more 3D models can be obtained according to application requirements.

[0040] For another example, in the business field invol...

Embodiment 2

[0068] This embodiment provides a training method for an image recognition model, such as Figure 4 shown, including:

[0069] Step 210, determine the category of the image to be recognized, and there are multiple categories;

[0070] Similar to Embodiment 1, the categories of images to be recognized in this embodiment can be defined according to the needs of different services.

[0071] Step 220, acquiring 3D models of the multiple categories, rendering the 3D models to generate images of the multiple categories, and obtaining an image set of a predetermined scale by changing rendering parameters;

[0072] The generation of the image set in this embodiment may adopt any of the methods described in Embodiment 1.

[0073] Step 230, using the image set to train an image recognition model.

[0074] In this embodiment, the image recognition model is a deep learning model used for image recognition. The deep learning model can be designed with reference to LeNet, AlexNet, VGG, ...

Embodiment 3

[0083] At present, after users see the decoration they like, they don't necessarily know the style of the decoration. They need to search a large amount of information on the Internet, and then select useful information. This embodiment applies the above image set generation method and image recognition model training method to the decoration business, which can solve this problem. Users only need to upload the corresponding images to obtain the required information.

[0084] The image set generation based on the present embodiment and the training method of the image recognition model include:

[0085]Step 1, determining the category of the image to be recognized, the category is set according to the decoration style of the architectural scene;

[0086] For example, for architectural scenes such as living room, bedroom, bathroom, dining room, balcony, etc., each scene defines multiple categories. For example, living room and bedroom can define 9 categories, corresponding to 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 invention provides an image set generating method and device and an image recognition model training method and system. A 3D model acquisition module acquires 3D models of multiple categories; animage set generating module is used for rendering the 3D models to generate images of the multiple categories, and image set with a preset scale is obtained through the change of rendering parameters.The image set can be used for training an image recognition model. The image set is generated through the 3D models and a rendering mode, and the generation efficiency of the image set is greatly improved. The data quality and training effects are good.

Description

technical field [0001] The present invention relates to image recognition, and more specifically, to a method and device for generating an image set, and a method and system for training an image recognition model. Background technique [0002] Image recognition refers to the technology of using computers to process, analyze and understand images to identify targets and objects in various patterns. Image recognition is an important field of artificial intelligence. In order to compile computer programs that simulate human image recognition activities, different image recognition models have been proposed for processing such as image classification and target location. The quality of the image recognition model plays a vital role in the effect of image recognition. [0003] Machine learning (Machine Learning, ML) specializes in the study of how computers simulate or implement human learning behaviors to acquire new knowledge or skills, and reorganize existing knowledge stru...

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): G06T15/00G06T15/04G06T15/50G06K9/00G06K9/62
CPCG06T15/00G06T15/005G06T15/04G06T15/506G06V20/64G06F18/24G06F18/214
Inventor 赵永科
Owner ALIBABA GRP HLDG 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