A training method and device for identifying a model

Abstract

The invention provides a training method and device for identifying a model, wherein, the method comprises the following steps: acquiring a first training set and a second training set; based on the first training set, training the ANN model through a plurality of iterative processes to obtain a plurality of trained ANN models; determining a test accuracy rate of each trained ANN model by the second training set, wherein the test accuracy rate is an average accuracy rate of all characters tested by the ANN model; according to the test accuracy of each ANN model after training, selecting the model with the highest accuracy as the global optimal recognition model. Since the scheme sets up a training set for training, and a training set for determining the accuracy rate, and the model with the highest accuracy rate is finally selected as the most final model, thus solving the technical problem that the recognition accuracy rate of the existing trained recognition model is low, and achieving the technical effect of effectively improving the recognition rate of the recognition model obtained by training.

Description

technical field [0001] The invention relates to the technical field of machine recognition, in particular to a training method and device for a recognition model. Background technique [0002] When training license plate characters through ANN (artificial neural network), first specify the number of iterations and step size, then generate a corresponding ANN classifier, and finally, recognize and classify characters through the trained ANN classifier. [0003] However, the existing ANN model training has the problem of overfitting caused by excessive iterations or underfitting caused by poor iterations. Furthermore, in the recognition process, no matter how many iterations, even if it fits just right, the local The phenomenon of misclassification and low positive detection rate. [0004] Aiming at the problem of low recognition accuracy of existing recognition models, no effective solution has been proposed yet. Contents of the invention [0005] An embodiment of the pre...

AI Technical Summary

Problems solved by technology

[0003] However, the existing ANN model training has the problem of overfitting caused by excessive iterations or underfitting caused by poor iterations. Furthermore, in the recognition process, no matter how many iterations, even if it fits just right, the local The phenomenon of misclassification and low positive detection rate

[0004] Aiming at the problem of low recognition accuracy of existing recognition models, no effective solution has been proposed yet

Images