Neural Network for Image Definition Prediction via Automated Label Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for training neural networks to predict image definition require manual labeling of thousands or tens of thousands of images, leading to high labor costs, low efficiency, and uncertain accuracy due to subjective judgments.
Innovation Solution
A method that involves obtaining an image set with partial manually labeled definition labels, extracting definition features, obtaining and correcting definition labels for additional images using these features, and iteratively expanding the image samples to train a neural network for predicting image definition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual labeling of thousands or tens of thousands of images is performed to train the neural network, then the training data quantity is sufficient, but the labor cost increases and work efficiency decreases
Solution Approach 1:
The system uses the neural network itself to perform automatic labeling of images based on extracted definition features, making the system self-sufficient in data preparation. The neural network predicts definition labels for images without requiring manual intervention, thereby eliminating the need for labor-intensive manual labeling while maintaining sufficient training data quantity.
Solution Approach 2:
The patent replaces the mechanical manual labeling process with an automated computational system. Instead of humans manually assigning labels, the system automatically extracts definition features and uses the neural network to generate labels, substituting human labor with automated mechanical processes.
2Quantity of substance
If manual labeling is performed to obtain image samples, then the labeling process can be completed, but the labeling accuracy is difficult to guarantee due to subjective judgments
Solution Approach 1:
The patent replaces subjective human judgment with objective automated feature extraction and neural network-based labeling. The system extracts definition features using computational methods and uses the trained neural network to generate consistent, objective labels, eliminating the variability and subjectivity inherent in manual labeling processes.
Solution Approach 2:
The system incorporates feedback mechanisms where the neural network's predicted labels are used to expand the training dataset iteratively. The extracted definition features provide feedback that guides the labeling process, ensuring consistency and accuracy by continuously refining the training data based on objective feature measurements rather than subjective human judgment.
3Quantity of substance
If a large number of image samples are manually labeled to train the neural network, then the training can proceed, but the training cost increases
Solution Approach 1:
The neural network system performs self-service by automatically generating training labels through feature extraction and prediction, eliminating the need for expensive manual labeling services. The system uses its own capabilities to prepare training data, thereby reducing or eliminating external training costs while maintaining sufficient data quantity.
Solution Approach 2:
The patent replaces costly manual labeling operations with free automated computational processes. Instead of paying humans to label images, the system uses automated feature extraction and neural network prediction, substituting expensive human labor with low-cost or free computational resources.
Data Source
AI summary
The present application disclose a neural network of predicting image definition, a training method and a prediction method. The training method includes: obtaining an image set and definition labels of some images in the image set, thereby obtaining image samples with the definition labels and to-be-expanded images except for the image samples; and extracting definition features of at least some images in the image set, obtaining definition labels of at least some images in the to-be-expanded images according to the extracted definition features, correcting the definition labels of the at least some images in the to-be-expanded images to expand the image samples, and using the image samples to train the neural network of predicting image definition, thereby obtaining a trained neural network.