Image Analysis Method for Identifying Incorrect Inference Causes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current analysis techniques fail to specify image sections causing incorrect inference in image recognition processes with sufficient precision using convolutional neural networks (CNNs).
Innovation Solution
An analysis method that generates a refine image to maximize correct label scores by changing incorrect inference images, and narrows down image regions to identify sections causing incorrect inference through a score-based approach, utilizing techniques like Grad-CAM for attention mapping and back-propagation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current analysis techniques (activation maximization, Grad-CAM, GBP) are used to identify image sections causing incorrect inference, then the analysis process can be performed, but the precision of specifying image sections is insufficient
Solution Approach 1:
The patent divides the image into multiple regions and processes each region separately through the neural network. By segmenting the image and analyzing regional characteristics independently, the system can precisely identify which specific region causes incorrect inference, thereby improving measurement precision and reliability simultaneously
Solution Approach 2:
The patent applies different analysis methods to different regions based on their local characteristics. Each region is evaluated with consideration of its specific properties, allowing for more accurate identification of problematic regions while maintaining overall analysis reliability
Data Source
AI summary
An analysis method implemented by a computer includes: generating a refine image by changing an incorrect inference image such that a correct label score of inference is maximized, the incorrect inference image being an input image when an incorrect label is inferred in an image recognition process; and narrowing, based on a score of a label, a predetermined region to specify an image section that causes incorrect inference, the score of the label being inferred by inputting to an inferring process an image obtained by replacing the predetermined region in the incorrect inference image with the refine image.