Character recognition and recognition model training methods, devices and systems, and storage medium

Abstract

Embodiments of the invention provide a character recognition model training method, device and system, a character recognition method, device and system, and a storage medium. The character recognition model training method comprises the steps of acquiring at least one sample image and labeled data corresponding to the at least one sample image, wherein the labeled data corresponding to each sample image is used for indicating the position of an actual character region of the sample image and actual characters in the actual character region; for each sample image in the at least one sample image, inputting the sample image into a region positioning network in a character recognition model to determine the position of a predicted character region of the sample image; inputting image information corresponding to the position of the predicted character region into a character recognition network in the character recognition model, so as to identify predicted characters in the predicted character region; and training the character recognition model by at least using the labeled data corresponding to the at least one sample image and the predicted characters. The accumulative error of character recognition can be effectively avoided.

Description

technical field [0001] The present invention relates to the field of image recognition, and more particularly to a character recognition model training method, device and system, a character recognition method, device and system, and a storage medium. Background technique [0002] Character recognition is required in many fields. The following takes bank card number recognition as an example for illustration. In the business of finance, finance, insurance and e-commerce, etc., it is often necessary to enter the user's bank card number information. With the popularization of smart devices and the rapid development of artificial intelligence technology, the current input of bank card numbers in many scenarios has changed from manual input to automatic algorithm recognition. These automatic recognition algorithms greatly reduce the time required for bank card number entry Time-consuming, but also reduces the possibility of manual input errors. [0003] Existing bank card num...

AI Technical Summary

Problems solved by technology

This multi-step method often brings a certain cumulative error in each decomposition step, and this error cannot be made up in subsequent steps, so there will be a bottleneck in the recognition accuracy

On the other hand, the multi-step method is also more complicated to implement, and it is difficult to achieve better recognition results.

Images