Image Determination Device Feature Segmentation for Model Explainability

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In image determination systems, it is unclear which features of an image are focused on for determination results, especially when additional training of a training model is performed, making it difficult to explain and ensure the reliability of the model.

Innovation Solution

An image determination device that includes feature extractors to output specific feature data of an image, a determiner to provide determination results based on this data, and a training part that updates the determiner using training data, allowing for clear identification of focused features even during additional training.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If additional training of a training model is performed using newly obtained training data, then the determination accuracy is improved, but it becomes unclear what features of the image are focused on to obtain the determination result

Engineering Contradiction:
Improvedetermination accuracyVSAvoidfeature interpretability
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The training model is segmented into multiple feature extractors, each responsible for extracting specific features from input images. This segmentation allows the system to maintain clarity about which features are being focused on, even when additional training is performed. Each feature extractor can be independently trained and its contributions tracked, preserving feature interpretability while improving determination accuracy through comprehensive training.

Inventive Principle:
Principle #1Segmentation

2Reliability

If the training model is updated through additional training, then the reliability of the model is improved, but it becomes difficult to explain what features of the image are focused on

Engineering Contradiction:
Improvemodel reliabilityVSAvoidexplainability
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The system incorporates feedback mechanisms that track and record which features are extracted and how they contribute to determination results. This feedback loop maintains transparency about feature usage even as the model is updated through additional training. The feature extractors provide continuous information about their output contributions, enabling explanation of model decisions while maintaining reliability through ongoing training updates.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS11922319B2Image determination device, training method and non-transitory computer readable medium storing program
Publication Date: 2024.03.05 OMRON CORP
  • US11922319B2 patent drawing
  • US11922319B2 patent drawing
  • US11922319B2 patent drawing

AI summary

Provided is an image determination device. The image determination device is provided with: feature extractors which output, on the basis of an image to be examined, each piece of feature data indicating a specific feature of the image; a determiner which outputs, on the basis of the feature data output from the extractors, output data indicating the determination result pertaining to the image; and a training part which trains the determiner so as to output, output data indicating the label data associated with the training image on the basis of the feature data output when the training image is input to the extractors, wherein the training part further trains, by using new training data, the determiner so that the output data indicating the label data associated with the image is output by the determiner.