Appearance based dominant eye selector systems and methods for gaze estimation from passive imagery
AssigneeBlink O.G. Ltd.
InventorsOren Haimovitch-Yogev, Gilad Drozdov, Igor Nor, Nadav Yehonatan Arbel
Life of this patent
Estimating the term…
Abstract
The disclosure relates to systems, methods, and programs for implicitly determining the relation of the eyes associated with gaze inference, including salient features (e.g., ocular dominance) that are not visible within a digital frame, allowing for real-time determination of the user's dominant eye for the purpose of gaze estimation, and point-of-regard mapping onto a 2D plane, achieved through an end-to-end training of an eye selector agent with domain-expertise knowledge embedded in an unsupervised manner via a convolutional deep neural network training process.
Claims16
A method for determining at least one dominant eye of a user for the task of gaze estimation, the method comprising:
- 1) obtaining at least one frame of a digital image of a user;
- 2) detecting at least one of a face, an eye region, or a head pose within the at least one frame using at least one first convolutional neural network applied to at least one data matrix;
- and
- 3) determining the user's dominant eye based on an extracted portion of the at least one of a face, an eye region, or a head pose, wherein the extracted portion includes left and right eye patches of the digital image of the user;
- and wherein the user's dominant eye is determined by applying at least one second convolutional neural network to the left and right eye patches of the digital image of the user.
A system for determining at least one dominant eye of a user for the task of gaze estimation, the system comprising:
- 1) an imaging device for obtaining at least one frame of a digital image of a user;
- 2) a processor configured to detect at least one of a face, an eye region, or a head pose within the at least one frame using at least one first convolutional neural network applied to at least one data matrix;
- and
- 3) a processor configured to determine the user's dominant eye based on an extracted portion of the at least one of a face, an eye region, or a head pose, wherein the extracted portion includes left and right eye patches of the digital image of the user;
- and wherein the processor configured to determine the user's dominant eye is further configured to apply at least one second convolutional neural network to the left and right eye patches of the digital image of the user.
Dependencies are inferred from the claim text.