A method for automatic construction of image datasets based on deep neural network
A technology of deep neural network and image data set, which is applied in the field of automatic construction of image data set based on deep neural network, and can solve problems such as inapplicable multi-objective
Active Publication Date: 2021-12-17
杭州玖维数字科技有限公司
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] In order to overcome the problem that the existing technology is not suitable for multi-target and high-resolution image data set construction, the present invention provides an automatic image data set construction method based on a deep neural network that effectively solves the above problems
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0027] The present invention will be further described below with reference to the accompanying drawings.
[0028] Refer Figure 1 ~ 5 A automatic construction method based on an image data set based on a depth neural network, the method comprising the steps of:
[0029] The first step, get the picture, the process is as follows:
[0030] Step 1.1: Determine the keyword information you want to retrieve, make sure it has the necessary contacts with the category tag (referred to as the target category) in the image data set to be buffered, such as the image data in the image data, can be used from the Internet Men, woman, child, pedestrian image acquisition;
[0031] Step 1.2: Use the picture search engine's API to retrieve the keyword information to determine, get the related picture and save these pictures;
[0032] In the second step, the image cleaning, the process is as follows:
[0033] Step 2.1: Determine a pre-trained neural network, such as Mask-RCNN-RESNET-101 that can be 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
Login to View More Abstract
A method for automatically constructing an image data set based on a deep neural network, the method for automatically constructing an image data set includes the following steps: the first step, acquiring pictures: determining the keyword information to be retrieved according to the target category, and using computer automatic programs on the Internet The images containing keywords are retrieved and saved; the second step is image cleaning: by using or improving the currently popular deep neural network, the saved keyword images are connected with the categories in the transitional data set, and valuable images are screened out. Suggested region images; the third step, category-based data set construction: for the saved suggested region images, the integration operation is performed according to the affiliation of keywords, transition categories and target categories, and the target image dataset is output. The invention provides a method for automatically constructing an image data set based on a deep neural network, which can solve the problem of insufficient number of training samples and insufficient types in the field of image recognition.
Description
Technical field [0001] The present invention belongs to the field of computer application, and specifically, the method of automatic construction method based on depth neural network is used. Background technique [0002] Data play an increasingly important role in the training of neural networks, such as in computer visual tasks, with the increase of training data, the performance of the model will grow a logarithmic level, so it is recommended that the researchers in the computer visual field should not The underestimation of the scale of data sets and suggests that they have worked more efforts in building a larger data set. However, building a clean, fair, diversified data set is usually a huge challenge because it will consume a lot of labor and funding. [0003] A series of research literature on the application of weak label technology in recent years. The core thinking of this technology is not only from the Internet, but also requires some additional pretreatment to re-c...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06N3/044G06F18/241
Inventor 周乾伟陈禹行郭媛刘一波陶鹏胡海根李小薪陈胜勇
Owner 杭州玖维数字科技有限公司