AI Confidence-Based Zoom for UI Object Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI systems face limitations in identifying graphical objects in user interfaces due to image quality issues, which restrict the development of automated tests.
Innovation Solution
An AI process is used to determine a confidence score for graphical objects in a user interface, and if the score is below a threshold, the system automatically zooms in or out to capture a higher resolution image, improving the object's identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If AI process is used to identify graphical objects in user interface images, then automated testing capability is improved, but identification accuracy deteriorates due to image quality limitations
Solution Approach 1:
The system dynamically adjusts the zoom level of captured images based on the confidence score of AI identification. When identification confidence is low, the system automatically zooms in to capture a closer view of the graphical object, and when confidence is high, it may zoom out to capture more context. This dynamic adjustment of image scale allows the AI system to maintain high identification accuracy across various image quality conditions, enabling more reliable automated testing.
2Measurement precision
If zoomed in image is taken to improve graphical object identification, then confidence score increases, but processing time and system complexity increase
Solution Approach 1:
The system implements a feedback mechanism where the AI process evaluates the confidence score of graphical object identification, and based on this feedback, automatically determines whether to capture a zoomed-in image. If the confidence score falls below a threshold, the system triggers a zoom-in capture to improve identification accuracy. This closed-loop feedback control ensures that zoom operations are performed only when necessary, optimizing the balance between identification accuracy and system complexity.
3Reliability
If zoom operations are performed automatically, then identification reliability improves, but processing time increases
Solution Approach 1:
The system applies partial action by performing zoom operations selectively rather than universally. Instead of zooming in on all graphical objects, the AI process first evaluates each object and only triggers zoom-in captures for objects with low confidence scores. This partial application of the zoom operation reduces unnecessary processing time while maintaining high identification reliability for objects that actually require enhanced visualization.
Data Source
AI summary
An image of a user interface of a device is captured. A graphical object is identified in the image of the user interface. For example, a menu item or a text object is identified in the image of the user interface. An Artificial Intelligence (AI) process is used to determine a confidence score for the graphical object; the confidence score identifies a confidence of how likely a type of the graphical object can be identified. In response to the first confidence score not meeting a threshold, a zoomed in image of the graphical object is taken or a zoomed-out image of the graphical object is taken. The zoomed in image or the zoomed-out image is used to increase the confidence score to better identify the type of the graphical object.


