Eye-Tracking Glint Segmentation for Accurate Gaze Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gaze estimation in eye tracking systems suffers from insufficient glint detection accuracy, which hinders precise gaze estimation and subsequent applications.
Innovation Solution
A deep learning-based method and apparatus that processes single-channel sample eyeball images with glint, generates multi-channel label images, performs semantic segmentation using a neural network model, and iteratively optimizes the model to accurately detect glint centers and ordering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional glint detection methods are used in eye tracking, then the system structure remains simple, but the glint detection accuracy is insufficient
Solution Approach 1:
The patent replaces traditional mechanical/optical glint detection methods with a deep learning-based neural network system. The neural network model processes eyeball images to detect glint positions, substituting conventional image processing algorithms with an intelligent system that learns patterns from training data, thereby significantly improving detection accuracy while accepting increased system complexity
Solution Approach 2:
The patent transforms the glint detection problem into a semantic segmentation task by changing the parameter representation from direct coordinate detection to pixel-wise classification. The neural network outputs a segmentation map where each pixel is classified as glint or non-glint, enabling more precise detection through probabilistic reasoning rather than deterministic thresholding
2Measurement precision
If deep learning-based glint detection is implemented, then glint detection accuracy is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by training the neural network model offline on a large dataset of eyeball images with labeled glint positions. This pre-training phase captures complex patterns and relationships in the data, allowing the model to make accurate predictions during real-time operation without requiring extensive computational resources during inference, thus reducing processing time while maintaining high accuracy
3Measurement precision
If deep learning-based glint detection is implemented, then glint detection accuracy is improved, but the model complexity and training requirements increase
Solution Approach 1:
The patent extracts and isolates the specific task of glint detection from the overall eye tracking system. By focusing the neural network exclusively on detecting glint positions in eyeball images rather than performing comprehensive eye analysis, the model complexity is reduced while maintaining high accuracy for the specific detection task. The segmented approach allows each component to be optimized independently
Data Source
AI summary
This application provides a deep learning-based method and apparatus for detecting glint in eye tracking. The method includes: processing and storing data sets of a single-channel sample eyeball image with glint in a txt file; generating a first multi-channel label image corresponding to the single-channel sample eyeball image; performing, through a preliminary neural network model, semantic segmentation on the data set corresponding to the single-channel sample eyeball image to output a second multi-channel label image; determining a loss function based on the first multi-channel label image and the second multi-channel label image; iteratively optimizing the preliminary neural network model through the loss function to obtain a final neural network model; and processing a single-channel test eyeball image through the final neural network model, and performing inference to obtain a glint center and glint ordering of the single-channel test eyeball image.