Eye Position Detection Using Segmented One-Eye and Both-Eyes Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in accurately detecting the position of eyes in images, particularly when the image contains both eyes, as they struggle to apply techniques developed for single-eye images, leading to reduced authentication accuracy and throughput.
Innovation Solution
An information processing system that uses a combination of one-eye and both-eyes images for learning, employing features extraction and similarity maps to accurately detect eye positions, enabling improved iris authentication by utilizing a both-eye image acquisition unit and eye position detection unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If techniques developed for single-eye images are applied to both-eye images, then existing systems can process images, but eye position detection accuracy deteriorates
Solution Approach 1:
The patent segments the learning process into separate models: a first learning model trained on one-eye images and a second learning model trained on both-eye images. This segmentation allows each model to specialize in its specific input type, preventing the application of single-eye techniques to both-eye images and thereby resolving the accuracy deterioration problem.
Solution Approach 2:
The system dynamically selects which learning model to apply based on the input image type. When a one-eye image is detected, the first learning model is activated; when a both-eye image is detected, the second learning model is activated. This dynamic adaptation ensures that the appropriate model is always used, maintaining high detection accuracy across different image types.
2Device complexity
If existing single-eye image techniques are used for both-eye images, then system complexity is reduced, but authentication accuracy and throughput deteriorate
Solution Approach 1:
The authentication process is segmented into distinct stages: first detecting whether the image contains one eye or both eyes, then selecting the appropriate learning model accordingly. This segmentation enables the system to maintain simplicity in each individual processing path while achieving high overall reliability through proper model selection.
3Device complexity
If a unified learning model is used for both one-eye and both-eye images, then model complexity is reduced, but detection accuracy for both-eye images deteriorates
Solution Approach 1:
The patent divides the learning model into two separate models: a first learning model for one-eye images and a second learning model for both-eye images. This segmentation allows each model to be optimized for its specific input characteristics, achieving high detection accuracy without requiring an overly complex unified model.
Solution Approach 2:
Instead of creating one overly complex universal model, the system uses two simpler, specialized models. This partial action approach (training separate models for specific cases) is more effective than attempting to train a single comprehensive model, thereby achieving better accuracy with lower individual model complexity.
Data Source
AI summary
An information processing system includes: an acquisition unit that obtains a both-eye image, which is an image of a face containing both eyes, from a target; and a detection unit that detects an eye position of the target in the both-eye image on the basis of a result of learning that uses a one-eye image containing only one of the eyes and the both-eyes image. According to such an information processing system, the eye position of the target contained in the both-eye image can be detected with high accuracy.