Systems and methods for actively re-weighting a plurality of image sensors based on content

1. A system, comprising:

a plurality of image sensors that output content descriptive of one or more visual attributes of a user, wherein each of the plurality of image sensors are positioned at a unique angle facing the user;

one or more processors; and

one or more non-transitory memory modules communicatively coupled to the one or more processors and storing machine-readable instructions that, when executed, cause the one or more processors to perform at least the following:

receive the content from the plurality of image sensors;

assign respective weights to each of the plurality of image sensors based on the content, wherein the respective weights of the plurality of image sensors are actively re-weighted over a period of time;

merge the content of the plurality of image sensors together based on the respective weights to create a single representation of the content; and

transform the single representation of the content into a task output, wherein the task output is representative of one or more physical attributes of the user.

2. The system of claim 1 , wherein the machine-readable instructions further cause the one or more processors to:

in response to receiving the content from the plurality of image sensors, calculate a compressed version of the content generated by each of the plurality of image sensors separately by a feature extractor.

3. The system of claim 2 , wherein the compressed version of the content generated by each of the plurality of image sensors are in the same format as one another such that the content from one of the plurality of image sensors is combinable with the content of the remaining plurality of image sensors.

4. The system of claim 1 , wherein assigning the respective weights to each of the plurality of image sensors is based on an amount of the content that a discrete image sensor provides that is useable when determining the one or more physical attributes of the user.

5. The system of claim 1 , wherein merging the content of the plurality of image sensors together to create the single representation of the content is based on one of the following:

a merge by valid approach including setting a weight for a particular stream of content that is invalid to zero and a weight of a remaining plurality of streams of content are normalized such that their combined sum is one;

a merge by mean confidence approach that is identical to the merge by valid approach except that the weight of the remaining plurality of streams of content are further weighted by a confidence value; and

a merge by maximum confidence approach that is identical to the merge by mean confidence approach except that a sub-network features are merged using an element-wise maximum value rather than a mean value.

6. The system of claim 1 , wherein the machine-readable instructions further cause the one or more processors to:

transform the single representation of the content into the task output based on either a neural network or a kernel function.

7. The system of claim 1 , wherein the task output includes at least one of gaze direction, mental state, emotion, identity recognition, facial expressions, movement of a head of the user, and tasks involving the head.

8. The system of claim 1 , wherein the image sensors include two or more of: a camera, a red, green, and blue (RGB) light sensor, an infrared sensor, and a depth sensor.

9. The system of claim 1 , wherein the system is part of a vehicle, and wherein the machine-readable instructions further cause the one or more processors to:

autonomously control at least one vehicle system based on the task output.

10. A vehicle, comprising:

a plurality of image sensors that output content descriptive of one or more visual attributes of a user, wherein each of the plurality of image sensors are positioned at a unique angle facing the user;

one or more processors; and

one or more non-transitory memory modules communicatively coupled to the one or more processors and storing machine-readable instructions that, when executed, cause the one or more processors to perform at least the following:

receive the content from the plurality of image sensors;

assign respective weights to each of the plurality of image sensors based on the content, wherein the respective weights of the plurality of image sensors are actively re-weighted over a period of time;

merge the content of the plurality of image sensors together based on the respective weights to create a single representation of the content;

transform the single representation of the content into a task output, wherein the task output is representative of one or more physical attributes of the user; and

autonomously control at least one vehicle system based on the task output.

11. The vehicle of claim 10 , wherein the machine-readable instructions further cause the one or more processors to:

in response to receiving the content from the plurality of image sensors, calculate a compressed version of the content generated by each of the plurality of image sensors separately by a feature extractor.

12. The vehicle of claim 11 , wherein the compressed version of the content generated by each of the plurality of image sensors are in the same format as one another such that the content from one of the plurality of image sensors is combinable with the content of the remaining plurality of image sensors.

13. The vehicle of claim 10 , wherein assigning the respective weights to each of the plurality of image sensors is based on an amount of the content that a discrete image sensor provides that is useable when determining the one or more physical attributes of the user.

14. The vehicle of claim 10 , wherein merging the content of the plurality of image sensors together to create the single representation of the content is based on one of the following:

a merge by valid approach including setting a weight for a particular stream of content that is invalid to zero and a weight of a remaining plurality of streams of content are normalized such that their combined sum is one;

a merge by mean confidence approach that is identical to the merge by valid approach except that the weight of the remaining plurality of streams of content are further weighted by a confidence value; and

a merge by maximum confidence approach that is identical to the merge by mean confidence approach except that a sub-network features are merged using an element-wise maximum value rather than a mean value.

15. The vehicle of claim 10 , wherein the machine-readable instructions further cause the one or more processors to:

transform the single representation of the content into the task output based on either a neural network or a kernel function.

16. The vehicle of claim 10 , wherein the task output includes at least one of gaze direction, mental state, emotion, identity recognition, facial expressions, movement of a head of the user, and tasks involving the head.

17. The vehicle of claim 10 , wherein the image sensors include two or more of: a camera, a red, green, and blue (RGB) light sensor, an infrared sensor, and a depth sensor.

18. A method, comprising:

receiving, by a computer, content from a plurality of image sensors, wherein the plurality of image sensors output content descriptive of one or more visual attributes of a user and are positioned at a unique angle facing the user;

assigning respective weights to each of the plurality of image sensors based on the content, wherein the respective weights of the plurality of image sensors are actively re-weighted over a period of time;

merging the content of the plurality of image sensors together based on the respective weights to create a single representation of the content; and

transforming the single representation of the content into a task output, wherein the task output is representative of one or more physical attributes of the user.

19. The method of claim 18 , further comprising:

in response to receiving the content from the plurality of image sensors, calculating a compressed version of the content generated by each of the plurality of image sensors separately by a feature extractor.

20. The method of claim 18 , further comprising:

transforming the single representation of the content into the task output based on either a neural network or a kernel function.