CRNN-based telecommunication industry intelligent customer service image recognition method and system
A technology of intelligent customer service and image recognition, which is applied in the field of image recognition, can solve the problems that the intelligent customer service system does not have image processing and cannot well meet the various needs of users, so as to improve the accuracy of image recognition and meet the needs of multi-modality, The effect of improving usage efficiency
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0033] In the present invention, CRNN adopts a three-layer structure, namely CNN (convolutional layer), which uses deep CNN to extract features from the input image to obtain a feature map; RNN (circular layer), which uses bidirectional RNN (BLSTM) to perform feature sequence Predict, learn each feature vector in the sequence, and output the predicted label (true value) distribution;
[0034] CTC loss (transcription layer), using CTC loss, converts a series of label distributions obtained from the recurrent layer into the final label sequence, so that the features of the image can be learned, and the word order features in the image can be expressed.
[0035] The main difference between the recognition method of the CRNN-based intelligent customer service image recognition system and the traditional CNN-based image recognition is that the system uses CNN to extract features and then uses RNN to learn word order features, which is more accurate than CNN. Learning the semantic i...
Embodiment 2
[0062] The system for image recognition of a CRNN-based intelligent customer service in the telecommunications industry described in Embodiment 2 is matched with the method for image recognition of a CRNN-based intelligent customer service in the telecommunications industry described in Embodiment 1, including a data collection server and a central server , the communication between the data acquisition server and the central server through a network cable;
[0063] The data collection server is used to collect pictures of problems related to the telecommunications field;
[0064] Establish a feature extraction module, a time series feature extraction module and a CTC loss update module in the central server, and the feature extraction module is used to extract image features from problem pictures related to the telecommunications field;
[0065] The time series feature extraction module is used to express the word order feature of the bidirectional LSTM on the image feature; ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More