Neural Network Training Image Clustering by Loss Value
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for improving image quality using neural network models lack efficiency in training each model for specific performance, as they do not effectively cluster training images to optimize model training.
Innovation Solution
An electronic device is designed to train neural network models by obtaining loss values from multiple models for different training images, identifying the smallest loss values, and clustering training images into groups for each model, thereby optimizing the training process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If training images are not clustered into groups for different neural network models, then the training process is simpler, but each neural network model cannot achieve specialized performance for specific image quality improvement methods
Solution Approach 1:
The patent segments the training process by clustering training images into multiple groups based on their characteristics and loss values. Each cluster corresponds to a specific neural network model, allowing targeted training for different image quality improvement methods (e.g., denoising, super-resolution, deblurring). This segmentation enables each model to specialize in specific tasks while maintaining manageable training complexity through automated clustering algorithms.
2Productivity
If multiple neural network models are trained with the same training images, then the training process is more efficient, but the models cannot achieve unique specialized performance
Solution Approach 1:
The patent applies local quality by assigning different training image clusters to different neural network models based on their specific functions. Each model receives training images that are locally optimized for its particular image quality improvement task. The system evaluates loss values for each model-image combination and assigns images to the most suitable model, ensuring that each model develops specialized performance while the overall training process remains efficient through parallel processing.
3Manufacturing precision
If training images are clustered into groups for each neural network model, then each model achieves targeted performance, but the training process becomes more complex
Solution Approach 1:
The patent implements self-service by using automated loss value evaluation and clustering algorithms that automatically assign training images to appropriate neural network models without manual intervention. The system evaluates each training image against multiple models, determines the optimal assignment based on loss values, and organizes clusters autonomously. This self-service approach reduces the perceived complexity for users while maintaining high image enhancement quality through scientifically-driven clustering.
Data Source
AI summary
An electronic device includes memory storing instructions; and one or more processors configured to execute the instructions to obtain first loss values by inputting a first training image into neural network models; identify a smallest first loss value from among the first loss values; identify the first training image as being in a first training image group for a first neural network model corresponding to the smallest first loss value; obtain second loss values by inputting a second training image into the neural network models; identify a smallest second loss value from among the second loss values; identify the second training image as being in a second training image group for a second neural network model corresponding to the smallest second loss value; train the first neural network model based on the first training image group; and train the second neural network model based on the second training image group.