Tencent Technology (Shenzhen) Company Limited

Tencent Technology Company Limited

Xiang Bai

Cong Yao

Baoguang Shi

Feiyue Huang

Xiaowei Guo

Wei Chen

Tingting Liu

Juhong Wang

Pengyuan Lv

Disclosed are a training method and apparatus for a CNN model, which belong to the field of image recognition. The method comprises: performing a convolution operation, maximal pooling operation and horizontal pooling operation on training images, respectively, to obtain second feature images; determining feature vectors according to the second feature images; processing the feature vectors to obtain category probability vectors; according to the category probability vectors and an initial category, calculating a category error; based on the category error, adjusting model parameters; based on the adjusted model parameters, continuing the model parameters adjusting process, and using the model parameters when the number of iteration times reaches a pre-set number of times as the model parameters for the well-trained CNN model. After the convolution operation and maximal pooling operation on the training images on each level of convolution layer, a horizontal pooling operation is performed. Since the horizontal pooling operation can extract feature images identifying image horizontal direction features from the feature images, such that the well-trained CNN model can recognize an image of any size, thus expanding the applicable range of the well-trained CNN model in image recognition.

Training method and apparatus for convolutional neural network model

Embodiments of this disclosure belong to the field of computer technologies and disclose a method and an apparatus for recognizing a character string in an image. The method includes: according to image data of a pre-stored sample image block including a character string, and based on a convolutional neural network algorithm, constructing an expression of a probability set corresponding to a character string recognition result of the sample image block, the expression of the probability set being constituted by a to-be-determined parameter; based on a training target of maximizing a probability that the character string recognition result determined according to the probability set is the character string included in the pre-stored sample image block, training the plurality of to-be-determined parameters to obtain a training value for each to-be-determined parameter; and when a to-be-recognized target image block is obtained, determining, based on the convolutional neural network algorithm and the training values of the plurality of to-be-determined parameters, a target probability set corresponding to the target image block, and determining, according to the target probability set, a character string recognition result of the target image block. By means of the embodiments of this disclosure, correctness of recognizing a character string can be improved.

Method and apparatus for recognizing character string in image

Growing community of inventors

Shenzhen, China

Xiang Bai