CNN Operation Module Real-Time Parameter Acquisition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition processing methods using Convolutional Neural Networks (CNNs) face challenges with real-time parameter acquisition and configuration, leading to poor performance due to the large number of operation parameters required across different layers.
Innovation Solution
An image recognition processing method and apparatus that obtains and configures CNN parameters in real-time through a data transmission bus, utilizing a convolutional neural network operation module with parallel operation components for efficient parameter distribution and processing, enabling flexible configuration and improved real-time performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a multi-layer convolutional neural network operation is performed on original image data, then image recognition accuracy is improved, but the number of operation parameters increases significantly, resulting in poor real-time performance
Solution Approach 1:
The patent divides the CNN operation into multiple independent operation components (first operation component, second operation component, etc.), where each component handles a specific layer or set of layers. This segmentation allows parallel processing of different CNN layers, reducing the time required to acquire and process all parameters simultaneously, thus improving real-time performance while maintaining multi-layer network accuracy.
Solution Approach 2:
The patent introduces a temporal dimension by acquiring operation parameters at different time points during the image recognition process. Instead of requiring all parameters to be available simultaneously before processing, the system acquires parameters progressively in time, enabling real-time parameter acquisition without compromising the multi-layer network operation accuracy.
2Adaptability or versatility
If a large number of operation parameters are required for CNN operations, then network capability is enhanced, but parameter acquisition real-time performance deteriorates
Solution Approach 1:
The patent performs preliminary actions by acquiring operation parameters for earlier layers before processing later layers. The first operation component acquires parameters in advance, and subsequent components continue acquiring parameters as needed. This preliminary acquisition strategy ensures that parameters are ready when required without causing delays in the overall processing timeline.
Solution Approach 2:
The patent maintains continuous parameter acquisition across multiple operation components. Instead of stopping parameter acquisition after the first layer, the system continuously acquires parameters for subsequent layers while the image recognition process progresses. This continuous action ensures that all necessary parameters are obtained without interrupting the overall processing flow, maintaining real-time performance.
3Measurement precision
If CNN operation parameters are configured for different layers, then processing accuracy is improved, but configuration complexity increases
Solution Approach 1:
The patent segments the parameter configuration task across multiple operation components, where each component is responsible for configuring parameters specific to its layer. This segmentation distributes the configuration complexity across independent units, making the overall system more manageable and easier to implement while maintaining layer-specific parameter accuracy.
Data Source
AI summary
An image recognition method and apparatus. The method comprises: obtaining original image data, convolutional neural network configuration parameters, and convolutional neural network operation parameters from a data transfer bus, the original image data comprising M pieces of pixel data, and M being a positive integer (101); and performing convolutional neural network operation on the original image data by a convolutional neural network operation module according to the convolutional neural network configuration parameters and the convolutional neural network operation parameters (102), wherein the convolutional neural network operation module comprises a convolution operation unit, a batch processing operation unit, and an activation operation unit connected in sequence. The method improves the real timeliness of image recognition.