Display Assistance Apparatus for Deep Learning Detection Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The evaluation of detection results from learning models using deep learning is hindered by overlapping displays of rectangular frames, object identifiers, and accuracy scores, making it difficult to confirm the content of detection results.
Innovation Solution
A display assistance apparatus and method that acquires detection results, displays position information and certainty scores for a predetermined number of detection targets, and allows switching between detection targets based on user instructions, improving visibility and ease of evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If detection results are displayed in an overlapping manner to show multiple targets, then the quantity of information displayed increases, but the ease of confirming detection result content deteriorates
Solution Approach 1:
The patent segments the display of detection results by dividing them into individual frames, each showing one detection target at a time. This segmentation allows the system to display multiple detection results (improving quantity of information) while preventing their overlap and maintaining clarity (preserving ease of confirmation). The detection result display unit presents each target in a separate, non-overlapping frame.
Solution Approach 2:
The patent transitions from a two-dimensional overlapping display to a sequential display approach where detection results are shown one after another in time. This dimensional change from spatial overlap to temporal sequence allows multiple detection targets to be displayed without overlapping, improving both the quantity of information and the ease of confirmation.
2Loss of information
If all detection results are displayed simultaneously to provide complete information, then the completeness of information improves, but the difficulty of viewing and evaluating individual results increases
Solution Approach 1:
The patent segments the complete set of detection results into individual frames, each containing one detection target. This segmentation maintains information completeness by ensuring all targets are displayed while making individual results viewable and evaluable without the confusion of overlapping displays.
Solution Approach 2:
The patent employs periodic action by sequentially displaying detection results in a time-based manner. Each detection target is displayed in turn, creating a periodic sequence that allows complete information presentation while maintaining ease of viewing and evaluation for each individual result.
Data Source
AI summary
A display assistance apparatus includes: a detection result acquisition unit that acquires a result of detecting an image including a plurality of detection targets, and in which detection processing is performed; a display processing unit; and an instruction acquisition unit that acquires information indicating an instruction to the result, wherein the display processing unit sets a predetermined number of the detection targets, as a detection result display target, and causes to display position information indicating a position of the detection target within the image, and a certainty score in association with the image, and, when the instruction acquisition unit acquires an instruction to switch the detection target serving as the result display target, the display processing unit switches the detection result display target to another one within the image, and causes to display the position information and the score related to the detection target after the switching.