Text recognition model training method and related products

Abstract

The embodiment of the present application discloses a text recognition model training method and related products. The method includes: performing a first perturbation on each original training sample in the original training sample set to obtain a first perturbation sample set; Each original training sample in the sample set is subjected to a second perturbation to obtain a second perturbation sample set; the text recognition model is optimized and trained according to the original training sample set, the first perturbation sample set and the second perturbation set . The embodiments of the present application help to improve the robustness of the text recognition model.

Description

technical field [0001] This application relates to the technical field of artificial intelligence, in particular to a text recognition model training method and related products. Background technique [0002] With the development of artificial intelligence technology, artificial intelligence products are applied to various scenes of life. For example, SIRI voice, face recognition, text recognition, etc. At present, when applying artificial intelligence technology, the neural network model is first trained, and the trained neural network model is applied to artificial intelligence equipment to realize intelligent recognition. However, when training the neural network model, a small perturbation of the training data is enough to change the final prediction result of the model, resulting in poor robustness of the trained neural network model. In order to improve the stability of the neural network model, Usually, during the training process, disturbances are automatically add...

AI Technical Summary

Problems solved by technology

However, when training the neural network model, a slight perturbation of the training data is enough to change the final prediction result of the model, resulting in poor robustness of the trained neural network model. In order to improve the stability of the neural network model, Usually, during the training process, disturbances are automatically added to the training samples, and new training samples are constructed to improve the accuracy of the model. However, currently, disturbances are added to the training samples by manual labeling, which makes the whole process cumbersome. The number of new training samples is limited, and the robustness of the optimized trained neural network model is still poor

Images