Detection model training method and device, single character detection method and device, equipment and medium

A detection model and training method technology, applied in the computer field, can solve the problems of personal privacy data leakage of certificate data, high data labeling costs, etc., and achieve the effect of reducing data labeling costs

Pending Publication Date: 2021-12-07
BEIJING WODONG TIANJUN INFORMATION TECH CO LTD +1
View PDF0 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In the process of realizing the present invention, the inventor found that there are at least the following technical problems in the prior art: the detection method based on the object detector needs a large amount of training data marked with single-character coordinates, the cost of data labeling is high, and during the process of labeling Credential data can easily cause personal privacy data leakage

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
  • Detection model training method and device, single character detection method and device, equipment and medium
  • Detection model training method and device, single character detection method and device, equipment and medium
  • Detection model training method and device, single character detection method and device, equipment and medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0039] Figure 1a It is a flowchart of a text word detection model training method provided by Embodiment 1 of the present invention. This embodiment is applicable to the situation when the text word detection model is trained. The method can be executed by a text word detection model training device, which can be implemented in software and / or hardware, for example, the text word detection model training device can be configured in a computer device. Such as Figure 1a As shown, the method includes:

[0040] S110. Acquire sample image data, wherein the sample image data includes sample position data marking the position of a single character and sample quantity data including the number of marked characters.

[0041] In this embodiment, in order to reduce the cost of data labeling during model training, part of the sample images can be marked with their word positions to obtain the sample position data marked with the word positions, and the number of characters in part of t...

Embodiment 2

[0059] figure 2 It is a flow chart of a text word detection method provided by Embodiment 2 of the present invention. This embodiment is applicable to the situation of detecting a single-character text box in an image. The method may be executed by a text word detection device, and the text word detection device may be implemented in software and / or hardware, for example, the text word detection device may be configured in a computer device. Such as figure 2 As shown, the method includes:

[0060] S210. Acquire an image to be detected.

[0061] In this embodiment, the image to be detected may be a single-line text image generated according to the image to be recognized requiring text recognition. Optionally, the image to be recognized may be detected by a text detector, and the image to be recognized may be preprocessed to obtain the image to be detected. The text detector can be detectors such as An Efficient and Accurate Scene Text Detector (EAST), natural scene text ...

Embodiment 3

[0068] An embodiment of the present invention provides a weakly supervised text word detection method based on deep learning, and this embodiment provides a preferred embodiment on the basis of the foregoing embodiments. This embodiment mainly includes two parts: model training and model prediction.

[0069] Figure 3a It is a flow chart of model training of a method for detecting single words in text with weak supervision based on deep learning provided by Embodiment 3 of the present invention. Such as Figure 3a As shown, the method includes:

[0070] S310. Generate training data with word positions.

[0071] In this embodiment, the training data includes image text data with word positions and weakly supervised image text data without word positions. Optionally, the form of the image text data is unnatural scene data including a single-line text image.

[0072] Optionally, the generation of text data with a single word position can be: collect corpus in the form of tex...

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 embodiment of the invention discloses a detection model training method, a single character detection method and device, equipment and a medium. The text single character detection model training method comprises steps of sample image data being acquired, and the sample image data comprising sample position data for labeling single character positions and sample number data for labeling character numbers; generating basic training sample data based on the sample position data, and generating weak supervision training sample data based on the sample quantity data; and training a pre-constructed text word detection model by using the basic training sample data and the weak supervision training sample data to obtain a trained text word detection model. According to the text word detection model training method provided by the embodiment of the invention, labeling cost of the sample image data is reduced, and the text word detection model with accurate prediction is trained on the basis of reducing the data labeling cost.

Description

technical field [0001] The embodiments of the present invention relate to the field of computer technology, and in particular to a detection model training method, a single character detection method, a device, equipment and a medium. Background technique [0002] Optical symbol recognition technology uses computer vision technology to extract text in images and convert them into text that humans and computers can understand. The single-character recognition method is one of the optical symbol recognition methods. The single-character recognition method is to detect a text area in an image, segment a single character in the text area, and classify and recognize a single character. For the single character recognition method, the detection and segmentation of a single character is a necessary step before character recognition. The character detection method based on deep learning is one of the commonly used single character segmentation methods. Specifically: take a single ch...

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 & AuthorityApplications(China)
IPC IPC(8): G06N3/04G06N3/08G06K9/34G06K9/62
CPCG06N3/08G06N3/045G06F18/214
Inventor谢树雷赖荣凤梅涛
OwnerBEIJING WODONG TIANJUN INFORMATION TECH CO LTD