Deep-Learning Image Feature Overlay for Relearning Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems lack an efficient method for users to easily select images that require relearning by a machine learning model, leading to unnecessary data redundancy and inefficiency in the learning process.
Innovation Solution
An image processing device and method that utilizes deep learning to extract feature amounts at multiple layers, superimposing these on images for easy visual differentiation between learned and unlearned images, allowing users to select images for relearning based on feature amount positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the user manually checks and selects images for relearning, then the learning model can be updated with relevant images, but the process is time-consuming and inefficient
Solution Approach 1:
The patent replaces the manual mechanical checking process with an automated computer vision system that uses deep learning models to automatically evaluate and select images for relearning based on feature extraction and comparison, eliminating manual time consumption while maintaining selection quality
Solution Approach 2:
The system enables self-service by allowing the learning model to automatically identify which images need relearning through automated feature amount comparison between learned and new images, making the selection process autonomous without requiring user intervention
2Reliability
If all images are used for relearning, then the learning model becomes more comprehensive, but redundant data increases processing overhead and reduces efficiency
Solution Approach 1:
The patent extracts only the necessary subset of images for relearning by comparing feature amounts between learned and new images, separating useful learning data from redundant data, thereby maintaining learning completeness while eliminating processing overhead from unnecessary images
Solution Approach 2:
Instead of processing all images equally, the system applies partial action by selectively processing only those images that meet the relearning criteria based on feature amount comparison, avoiding excessive processing of redundant images while ensuring all necessary images are included
3Loss of information
If feature amounts from multiple layers are extracted and displayed, then the user can better understand image differences, but the display complexity and information processing increase
Solution Approach 1:
The patent segments the feature amount information from different deep learning layers and displays them separately and systematically, allowing users to understand features at different abstraction levels without presenting a overwhelming wall of combined information, thus reducing perceived complexity while maintaining information completeness
Data Source
AI summary
Provided is an image processing device that enables a user to easily select an image to be newly learned. Image processing device includes learning part, acquisition part, and display controller. Learning part learns an image by using deep learning. Acquisition part acquires a first feature amount of each of a first image and a second image extracted in a first layer in deep learning, and a second feature amount of each of a first image and a second image extracted in a second layer different from the first layer in deep learning. Display controller superimposes and displays first feature amount information on a first feature amount and a second feature amount extracted from a first image on the first image, and superimposes and displays second feature amount information on the first feature amount and the second feature amount extracted from the second image on the second image.