Rotated Line Character Training Sample Generation for Vertical Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition technologies cannot directly utilize horizontal text line images to train vertically-oriented text recognition models, resulting in the need for extensive collection of vertically-oriented text images, which is resource-intensive and costly.
Innovation Solution
Generate rotated line character training samples by rotating character units in horizontal text images by 90 degrees and use these samples to train neural networks for image character recognition models, allowing for the recognition of vertically-oriented characters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vertically-oriented text images are collected to train a vertically-oriented text recognition model, then the recognition model training performance is improved, but the manpower and material resources are wasted
Solution Approach 1:
The patent creates rotated copies of existing horizontal text line images to generate vertically-oriented training samples. By rotating the images 90 degrees, the system reuses existing horizontal text resources as vertically-oriented training data, eliminating the need to manually collect vertically-oriented text images while maintaining training effectiveness
Solution Approach 2:
The patent changes the orientation parameter of existing text images by rotating them 90 degrees. This transforms horizontal text line images into vertically-oriented images, allowing the system to generate vertically-oriented training samples from horizontal text resources through parameter transformation rather than manual collection
2Quantity of substance
If existing horizontal text line images are used to train a vertically-oriented text recognition model, then the resource consumption is reduced, but the training effectiveness deteriorates
Solution Approach 1:
The patent performs preliminary rotation of horizontal text line images to create vertically-oriented training samples before training the recognition model. This preprocessing step ensures that the training data matches the target recognition task, maintaining training effectiveness while reducing resource consumption by reusing existing horizontal text resources
Solution Approach 2:
The patent creates rotated copies of existing horizontal text images to generate vertically-oriented training samples. This copying approach with rotation transformation allows the system to maintain training effectiveness by providing appropriately oriented samples while conserving resources by reusing existing horizontal text image databases
Data Source
AI summary
Embodiments of the present disclosure disclose an image character recognition model generation method and apparatus, and a vertically-oriented character image recognition method and apparatus. The image character recognition model generation method includes: generating a rotated line character training sample, wherein the rotated line character training sample includes a rotated line character image and an expected character recognition result corresponding to the rotated line character image, and there is a difference of 90 degrees between character units in the rotated line character image and character units in a standard line character image; and training a set neural network by using the rotated line character training sample, to generate an image character recognition model. The technical solutions of the embodiments of the present disclosure overcome the technical defect that existing line character images cannot be directly used to train a vertically-oriented character image recognition model, thereby implementing the efficient recognition of vertically-oriented characters.