Object Classification Using Prediction Consistency for Class Imbalance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image object classification models face challenges due to class imbalance, where unequal training data distribution leads to degraded classification performance, particularly for classes with insufficient data.
Innovation Solution
A computing device uses a prediction consistency value to classify objects as majority or minority classes based on the consistency of class prediction values, employing suitable class predictors for each class type to improve classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single classification model is used for all classes, then the device complexity is reduced, but the classification accuracy for minority classes deteriorates due to class imbalance
Solution Approach 1:
The patent divides the classification task into two separate models: a majority class prediction model and a minority class prediction model. This segmentation allows each model to be optimized for its specific class type, improving classification accuracy for minority classes without significantly increasing overall system complexity.
Solution Approach 2:
The patent dynamically selects which prediction model to use based on the input data characteristics. The system determines whether the input belongs to a majority or minority class and accordingly selects the appropriate prediction model, making the classification system adaptive rather than static.
2Measurement precision
If additional training data for minority classes is collected, then the classification accuracy for minority classes is improved, but the time and resources required for data collection and training increase
Solution Approach 1:
The patent performs preliminary actions during the training phase by creating separate prediction models for majority and minority classes. This preliminary differentiation allows the system to handle class imbalance without requiring additional minority class training data, saving time and resources that would otherwise be needed for data collection and retraining.
Solution Approach 2:
The patent changes the parameter of model architecture by creating distinct prediction models with different structures optimized for majority versus minority classes. This parameter change enables the system to achieve better minority class accuracy without increasing training data requirements.
3Measurement precision
If multiple prediction models are maintained for different class types, then the classification accuracy is improved, but the device complexity and computational resources increase
Solution Approach 1:
The patent segments the prediction models into two distinct types: majority class prediction models and minority class prediction models. This segmentation organizes the complexity in a manageable way, allowing the system to maintain multiple models without overwhelming computational overhead.
Solution Approach 2:
The patent applies local quality by designing prediction models with structures specifically optimized for their intended class type. Majority class models and minority class models have different architectural characteristics suited to their respective tasks, improving overall accuracy without unnecessary complexity.
4Reliability
If ensemble models are used for majority class prediction, then the classification robustness is improved, but the computational resources and processing time increase
Solution Approach 1:
The patent dynamically selects between using an ensemble of majority class models or a single target classification model based on the specific prediction task. This dynamic approach allows the system to achieve robustness when needed while conserving computational resources when a single well-trained model suffices.
Data Source
AI summary
Disclosed is a method that includes generating a prediction consistency value that indicates a consistency of prediction of an object in an input image with respect to class prediction values for the object in an input image from classification models to which the input image is input, and identifying a class of the object. Identifying the class of the object includes, in response to a class type being determined, based on the prediction consistency value, of the object being determined to correspond to a majority class, identifying a class of the object based on a corresponding class prediction value output for the object from a majority class prediction model, and in response to the class type of the object being determined to correspond to a minority class, identifying the class of the object based on another corresponding class prediction value output for the object from a minority class prediction model.