Online correction for context-aware image analysis for object classification
A computer system records eye tracking data and identifies movements in the eye tracking data to create and iteratively refine associations between eye features and basic geometric shapes. The associations are used to rapidly identify eye features for eye tracking. The system derives performance metrics and fatigue estimations from the identified basic geometric shapes, including changes in such shapes over time. The system continuously adjusts and/or weights the associations in real-time based on eye tracking data.
1 . A computer apparatus comprising:
at least one eye tracking camera; and
at least one processor in data communication with a memory storing processor executable code,
wherein the processor executable code configures the at least one processor to:
receive an image stream from the at least one eye tracking camera;
identify a plurality of basic geometric shapes in the image stream; and
generate a pose estimate from the plurality of basic geometric shapes.
2 . The computer apparatus of claim 1 , wherein the processor executable code further configures the at least one processor to generate a fatigue estimate based on the pose estimate.
3 . The computer apparatus of claim 2 , wherein:
the processor executable code further configures the at least one processor to receive task specific data; and
generating the fatigue estimate is based on the task specific data.
4 . The computer apparatus of claim 3 , wherein the at least one processor is configured as a trained neural network configured to receive the basic geometric shapes and task specific data as inputs.
5 . The computer apparatus of claim 1 , wherein the at least one eye tracking camera comprises a neuromorphic event camera.
6 . The computer apparatus of claim 5 , wherein the basic geometric shapes are defined by at least an interface between eye features.
7 . The computer apparatus of claim 1 , wherein the processor executable code further configures the at least one processor to record the image stream and basic geometric shapes in a training set of data for a machine learning algorithm.
8 . A method comprising:
receiving an image stream from at least one eye tracking camera;
identifying a plurality of basic geometric shapes in the image stream; and
generating a pose estimate from the plurality of basic geometric shapes.
9 . The method of claim 8 , further comprising generating a fatigue estimate based on the pose estimate.
10 . The method of claim 9 , further comprising receiving task specific data, wherein generating the fatigue estimate is based on the task specific data.
11 . The method of claim 10 , further comprising instantiating a trained neural network configured to receive the basic geometric shapes and task specific data as inputs.
12 . The method of claim 8 , wherein the at least one eye tracking camera comprises a neuromorphic event camera.
13 . The method of claim 12 , wherein the basic geometric shapes are defined by at least an interface between eye features.
14 . The method of claim 8 , further comprising recording the image stream and basic geometric shapes in a training set of data for a machine learning algorithm.
15 . A pilot fatigue monitoring system comprising:
at least one eye tracking camera; and
at least one processor in data communication with a memory storing processor executable code,
wherein the processor executable code configures the at least one processor to:
receive an image stream from the at least one eye tracking camera;
identify a plurality of basic geometric shapes in the image stream;
generate a pose estimate from the plurality of basic geometric shapes; and
generate a fatigue estimate based on the pose estimate.
16 . The pilot fatigue monitoring system of claim 15 , wherein:
the processor executable code further configures the at least one processor to receive task specific data; and
generating the fatigue estimate is based on the task specific data.
17 . The pilot fatigue monitoring system of claim 16 , wherein the at least one processor is configured as a trained neural network configured to receive the basic geometric shapes and task specific data as inputs.
18 . The pilot fatigue monitoring system of claim 15 , wherein the at least one eye tracking camera comprises a neuromorphic event camera.
19 . The pilot fatigue monitoring system of claim 18 , wherein the basic geometric shapes are defined by at least an interface between eye features.
20 . The pilot fatigue monitoring system of claim 15 , wherein the processor executable code further configures the at least one processor to record the image stream and basic geometric shapes in a training set of data for a machine learning algorithm.