Scene Text Detection Method of Arbitrary Shape Based on Margin Constraint
A text detection, arbitrary shape technology, applied in the field of artificial intelligence, can solve the problem that the constraint information is not well used and so on
Active Publication Date: 2021-02-26
EAST CHINA NORMAL UNIV
View PDF11 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
The progressive expansion of the network does not make good use of the constraint information existing between each text instance
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
[0018] First, the deep convolutional neural network obtains the deep feature information of the text image to be detected. Inspired by the attention mechanism, the attention mechanism is used to introduce the global context information as prior knowledge into the channel selection. Multi-scale information is integrated, the global information provided by global pooling is used as a guide to select the underlying features, channel processing is performed on low-level features, and then the information after global pooling is used for weighting to obtain weighted underlying features, and then upsampled. Then it is added to the high-level semantic feature information to obtain the deep feature information.
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
The invention discloses a scene text detection method of arbitrary shape based on margin constraints. The method extracts features in a scene picture through a deep network, processes the feature information to generate a text instance graph to complete the text detection. At the same time, we make full use of the margin constraints existing between text instances, and use the generated margin as an auxiliary means of the entire network to correct the text instance graph and produce more accurate detection results. The present invention performs text detection on a given natural scene picture, and realizes high-precision output of the coordinate position of the corresponding text area.
Description
technical field [0001] The invention belongs to a method for text detection in the field of artificial intelligence, relates to the technical fields of computer vision and machine learning, and in particular relates to a method for detecting text in scenes of arbitrary shapes based on margin constraints. Background technique [0002] In the Internet world, pictures are an important medium for conveying information. Especially in e-commerce, social networking, search and other fields, hundreds of trillions of images are transmitted every day. Text in natural scene pictures is a hot task full of challenges. Different from document text recognition, text recognition in natural scenes has problems such as complex image background, low resolution, various fonts, and different shapes. Traditional optical character recognition is here. Not applicable in such cases. In order to better perform natural scene text recognition, it is necessary to achieve more accurate detection of sce...
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): G06K9/32G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/62G06V30/10G06N3/045G06F18/241
Inventor 马天龙李鑫
Owner EAST CHINA NORMAL UNIV



