Grouped Convolutional Networks for Mobile Character Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition technologies on mobile devices face a trade-off between reducing calculation resources and maintaining recognition accuracy, leading to poor overall recognition performance due to the reduction of convolutional neural network parameters.
Innovation Solution
Implement a grouped convolutional neural network model that divides input data into groups, using a combination of standard and group convolutions with reduced parameters and increased network depth to process multiple groups simultaneously, enhancing recognition accuracy while minimizing calculation load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the quantity of network parameters at each layer of the convolutional neural network model is reduced, then the calculation amount is reduced and recognition speed is improved, but the recognition accuracy is reduced
Solution Approach 1:
The patent applies segmentation by dividing the input image into multiple grid cells and processing each cell independently through separate convolutional neural network sub-models. This segmentation allows the system to reduce the number of parameters in each sub-model while maintaining overall recognition accuracy through parallel processing of multiple segments.
Solution Approach 2:
The patent introduces a spatial dimension by gridting the input image into multiple cells and creating multiple parallel sub-models, each processing a specific spatial region. This dimensional transformation from a single large model to multiple smaller models arranged in a spatial grid enables parameter reduction while preserving accuracy.
2Power
If the quantity of network parameters is reduced, then the calculation amount is reduced, but the recognition effect becomes poor
Solution Approach 1:
The patent segments the recognition task into multiple independent sub-tasks, each handled by a separate sub-model processing a specific grid cell. This segmentation reduces the parameter count in each sub-model while the collective output of all sub-models maintains high recognition reliability.
Solution Approach 2:
The patent merges the outputs of multiple sub-models by aggregating predictions from all grid cell processors. This combining mechanism allows the system to achieve high recognition reliability through ensemble voting while each individual sub-model operates with reduced parameters.
Data Source
AI summary
This application discloses a character recognition method and apparatus, an electronic device, and a medium. The character recognition method includes: obtaining a character picture, where the character picture includes at least one character; inputting the character picture to a grouped convolutional neural network model for prediction, to obtain character sequence prediction information corresponding to the character picture; and obtaining, based on the character sequence prediction information, a character recognition result corresponding to the character picture.