Text line recognition method and device, readable storage medium and electronic equipment

Abstract

The invention relates to a text line recognition method and device, a readable storage medium and electronic equipment. The text line recognition method comprises the steps that at least one to-be-detected text line image is input into a preset neural network model, a detection result output by the preset neural network model is obtained, and the detection result is a character string in the to-be-detected text line image, wherein the preset neural network model is obtained by training by taking the synthesized image sample as a training sample. A composite image sample comprises a sample, a positive sample, a difficult negative sample and a common negative sample. According to the technical scheme, a synthetic image sample comprising a sample, a positive sample, a difficult negative sample and a common negative sample is used as a training sample of a preset neural network, the training sample of the preset neural network is enabled to be the associated image data, so that more supervision information is introduced, and the training effect of the preset neural network model is improved in comparison with the mode of adopting mutually independent random text line images as the training data for training.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a text line recognition method and device, a readable storage medium, and electronic equipment. Background technique [0002] Text line recognition, that is, input the text line image into the trained neural network model, and obtain a string output. The existing method for training a neural network model for text line recognition uses independent random text line images as training data for training, and does not consider the training data sampling strategy, resulting in poor model training effect. Contents of the invention [0003] The object of the present invention is to provide a text line recognition method and device, a readable storage medium, and an electronic device, by using synthetic image samples including samples, positive samples, difficult negative samples, and common negative samples as training samples for a preset neural network , introduce more supe...

AI Technical Summary

Problems solved by technology

The existing method of training the neural network model for text line recognition is to use independent random text line images as training data for training, and does not consider the training data sampling strategy, resulting in poor model training effect

Images