3D Pose-Based Gaze Direction Estimation From 2D Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating gaze direction in two-dimensional camera images suffer from low accuracy due to poor estimation of three-dimensional pose and coordinates, leading to incorrect determination of the person's focus point.
Innovation Solution
A device and method utilizing a neural network model that inputs three-dimensional pose and position information from a two-dimensional image to estimate the gaze direction, employing techniques like supervised learning and neural networks to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If three-dimensional pose and coordinates estimation is used to determine gaze direction, then the method can be implemented with existing computer vision techniques, but the estimation accuracy becomes low leading to incorrect gaze direction determination
Solution Approach 1:
The patent replaces traditional geometric and mechanical pose estimation methods with a neural network-based deep learning system. The neural network model learns complex patterns from training data to directly predict accurate gaze direction, substituting the mechanical 3D pose estimation approach with an intelligent system that achieves superior accuracy without relying on precise 3D coordinate reconstruction.
2Adaptability or versatility
If traditional three-dimensional pose estimation methods are used, then the system can process two-dimensional images, but the front direction of the face region cannot be correctly estimated
Solution Approach 1:
The patent changes the fundamental parameters used for gaze estimation by training the neural network to directly predict gaze direction angles (azimuth and elevation) rather than relying on intermediate 3D pose parameters. This parameter transformation allows the system to maintain adaptability for processing 2D images while achieving reliable and accurate gaze direction estimation through learned patterns from training data.
Data Source
AI summary
Estimate processing is performed to estimate a gaze direction of a person shown in a two-dimensional image. In the estimate processing, three-dimensional pose information on a target person is acquired from the two-dimensional image in which the target person of the estimation of the gaze direction is shown. In the estimate processing, three-dimensional position information on an object shown in the two-dimensional image is acquired from the two-dimensional image. In the estimate processing, input information is further input to a neural network model that outputs the gaze direction of the person, and the output information on the neural network model is acquired as the gaze direction of the target person. The input information on the neural network includes the three-dimensional pose information and the three-dimensional position information acquired by the estimate processing.