Character recognition method, device and equipment and readable storage medium

Abstract

The embodiment of the invention discloses a character recognition method, device and equipment and a readable storage medium. The method comprises the steps of cutting an image with characters to be recognized into a plurality of character strips according to the arrangement direction of the characters, obtaining feature maps of the character strips by utilizing a full convolutional network recognition model, performing deformation processing on the feature maps of the character strips to obtain a feature sequence, carrying context information, of the character strips, and predicting the characters in the character strips based on the feature sequence. The convolution operation does not depend on the state of the previous step and is irrelevant to the length of the input sequence, so thatparallel computing can be performed, the modeling process of the feature sequence is greatly accelerated, and compared with a recurrent neural network, the convolution network has the advantages of fewer parameters, lower computing complexity, less occupied memory space and running time and easiness in deployment.

Description

technical field [0001] The present application relates to the technical field of information processing, and more specifically, to a character recognition method, device, equipment and readable storage medium. Background technique [0002] Optical Character Recognition (OCR) is a sub-direction of computer vision, the goal is to recognize text from image data and save it as computer text data. [0003] At present, recognition models based on recurrent neural networks (such as long-term short-term memory network LSTM, or bidirectional LSTM) are mostly used for character recognition, but recognition models based on recurrent neural networks are difficult to converge during training, and it is also difficult to perform parallel computing. Moreover, due to the relatively large amount of data and computation, the actual deployment process in the later stage is also relatively difficult. In addition, the recognition accuracy of the current character recognition method is low, and ...

AI Technical Summary

Problems solved by technology

[0003] At present, recognition models based on recurrent neural networks (such as long-term short-term memory network LSTM, or bidirectional LSTM) are mostly used for character recognition, but recognition models based on recurrent neural networks are difficult to converge during training, and it is also difficult to perform parallel computing. Moreover, due to the relatively large amount of data and computation, the actual deployment process in the later stage is also relatively difficult

In addition, the recognition accuracy of the current character recognition method is low, and problems such as multiple characters, missing characters, and typos are prone to occur.

Images