Automated Deep Learning Network Selection for Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Selecting a suitable Convolutional Neural Network (CNN) algorithm for image processing is challenging for developers without extensive experience, as it requires deep professional knowledge and numerous analysis experiments.
Innovation Solution
A device and method for selecting a deep learning network that includes modules for obtaining the problem type, selecting and dividing data sets, calculating similarity using algorithms like SSIM, adjusting batch sizes, training, and testing deep learning networks to identify the optimal algorithm for specific image processing tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a developer selects a suitable CNN algorithm from multiple CNN algorithms, then the image processing model can be built with appropriate performance, but the developer needs rich development experience and deep professional background
Solution Approach 1:
The patent introduces an automated selection system that acts as an intermediary between the developer and the complex task of selecting CNN algorithms. The system automatically evaluates multiple CNN algorithms against the specific image processing task requirements, eliminating the need for developers to have deep professional expertise while still achieving reliable model performance.
Solution Approach 2:
The system performs self-service by automatically analyzing the image processing task, evaluating multiple CNN algorithms, and selecting the most suitable one without requiring extensive developer intervention or expertise. The automated evaluation process independently determines the optimal algorithm based on task characteristics.
2Measurement precision
If a developer conducts a large number of analysis experiments to select the most suitable CNN algorithm, then the selection accuracy improves, but the time and resources required increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-establishing an automated evaluation framework that assesses CNN algorithms based on task characteristics before actual model training begins. The system performs preliminary analysis of task requirements and algorithm suitability, eliminating the need for extensive time-consuming experiments during the development phase.
Solution Approach 2:
The patent replaces the mechanical process of conducting numerous manual analysis experiments with an automated computational evaluation system. The system uses algorithmic assessment methods to determine CNN suitability, substituting time-consuming manual experimentation with efficient automated processing that achieves comparable or superior selection accuracy.
Data Source
AI summary
A method for selecting a deep learning network which is optimal for solving an image processing task obtaining a type of the image processing task, selecting a data set according to the type of problem, and dividing selected data set into training data and test data. Similarities between different training data are calculated, and a batch size of the training data is adjusted according to the similarities of the training data. A plurality of deep learning networks is selected according to the type of problem, and the plurality of deep learning networks is trained through the training data to obtain network models. Each of the network models is tested through the test data, and the optimal deep learning network with the best test result is selected from the plurality of deep learning networks appropriate for image processing.