Correction Learning Model for Inference Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model learning methods are limited in achieving high accuracy inference due to the likelihood of missing hidden feature points, and they are often specialized for specific causes of accuracy deterioration, making them ineffective when the cause is unclear.
Innovation Solution
A learning device and method that generates probabilistic inference results, formats them for correction, and trains a correction learning model using input data, correct answer data, and formatted inference results to improve inference accuracy without relying on the specific cause of deterioration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a specialized method is used to extract feature points and compensate for missing feature points, then accuracy is improved for specific cases, but the method cannot handle various causes of accuracy deterioration
Solution Approach 1:
The patent creates a universal correction learning model that can handle various causes of accuracy deterioration (occlusion, blur, low resolution, etc.) through a single unified approach. The model takes as input the original image, deteriorated image, and deterioration type, and outputs corrected feature points regardless of the specific deterioration cause, making the system adaptable to multiple scenarios rather than requiring specialized methods for each case
Solution Approach 2:
The patent changes the parameters of the learning model by introducing a correction learning model with specific loss functions that account for different deterioration types. The model uses parameter adjustments in the loss function (e.g., weighting different error terms based on deterioration type) to adapt to various accuracy deterioration causes while maintaining a single unified model structure
2Measurement precision
If a correction learning model is trained to correct formatted inference results, then inference accuracy is improved, but training complexity and computational resources increase
Solution Approach 1:
The patent performs preliminary formatting of the inference results before feeding them into the correction learning model. By pre-processing the probabilistic inference results into a standardized format and pre-selecting appropriate training data with known deterioration types, the system reduces the complexity of the actual training process and makes the training more efficient and manageable
Data Source
AI summary
A learning device 1X includes a probabilistic inference result generation means 16X, a formatting means 17X, and a training means 18X. The probabilistic inference result generation means 16X is configured to generate a probabilistic inference result that is probabilistically generated for an input data. The formatting means 17X is configured to generate a formatted inference result obtained by formatting the probabilistic inference result. The training means 18X is configured to train a correction learning model that is a learning model configured to correct the formatted inference result, based on the input data, correct answer data corresponding to the input data, and the formatted inference result.