Sensor system based on stacked sensor layers

1. A sensor assembly comprising:

a sensor layer, of a plurality of stacked layers, including an array of pixels configured to capture one or more first images; and

one or more processing layers of the plurality of stacked layers located in the sensor assembly beneath the sensor layer, the one or more processing layers configured to:

extract features of one or more objects and their pixel locations from the one or more first images; and

enable a subset of the array of pixels based on the pixel locations to capture one or more second images of the one or more objects.

2. The sensor assembly of claim 1 , wherein the one or more processing layers comprise an analog to digital conversion (ADC) layer with logic for conversion of analog intensity values captured by the array of pixels into digital values.

3. The sensor assembly of claim 2 , further comprising an interface connection between each pixel in the array of pixels and the logic of the ADC layer.

4. The sensor assembly of claim 1 , wherein the sensor layer comprises a non-silicon photodetector material.

5. The sensor assembly of claim 4 , wherein the non-silicon photodetector material is selected from a group consisting of a Quantum Dot (QD) photodetector material and an organic photonic film (OPF) photodetector material.

6. The sensor assembly of claim 1 , wherein at least one of the one or more processing layers includes logic configured to determine depth information for the one or more objects based in part on the extracted features.

7. The sensor assembly of claim 1 , wherein at least one of the one or more processing layers includes a convolutional neural network (CNN).

8. The sensor assembly of claim 7 , wherein a plurality of network weights in the CNN are trained for at least one of classification of the captured one or more first images and recognizing at least one feature in the captured one or more first images.

9. The sensor assembly of claim 8 , wherein the trained network weights are applied on data associated with an image of the captured one or more first images to determine a classifier of the image.

10. The sensor assembly of claim 7 , wherein the CNN includes an array of memristors configured for storage of the trained network weights.

11. The sensor assembly of claim 1 , further comprising:

an optical assembly positioned on top of the sensor layer configured to direct the reflected light toward the array of pixels.

12. The sensor assembly of claim 11 , wherein the optical assembly includes one or more layers of wafers stacked on top of the sensor layer, each wafer of the optical assembly being implemented as a glass wafer and configured as an individual optical element of the optical assembly.

13. The sensor assembly of claim 1 , wherein the one or more processing layers are configured to:

transmit information related to the pixel locations of the features of the one or more objects in the one or more first images to a host device;

receive, from the host device, a map of pixel locations based on the information; and

enable the subset of the array of pixels based on the map of pixel locations to capture the one or more second images.

14. The sensor assembly of claim 13 , wherein the one or more processing layers are configured to:

generate a key frame including the information and pixel locations of features of a second object different from the one or more objects; and

transmit the key frame to the host device.

15. The sensor assembly of 14 , wherein the map of pixel locations is a first map of pixel locations; and

wherein the one or more processing layers are configured to:

generate, based on one or more second images, a second map of pixel locations of the features of the one or more objects; and

transmit the second map to the host device.

16. The sensor assembly of claim 15 , wherein each of the second map and the key frame is transmitted as part of a frame; and

wherein the second map is transmitted at a higher frame rate than the key frame.

17. The sensor assembly of claim 7 , wherein the CNN is configured to generate filtered outputs of the one or more first images; and

wherein the one or more processing layers comprise feature extraction circuits to extract the features from the filtered outputs.

18. The sensor assembly of claim 17 , wherein the feature extraction circuits are configured to perform at least one of: extracting dots based on a centroid estimation operation, or extracting keys or events based on a threshold detection operation.

19. The sensor assembly of claim 3 wherein the one or more processing layers comprise a first processing layer and a second processing layer;

wherein the interface connection between the sensor layer and the first processing layer comprises copper pad connections; and

wherein the first processing layer and the second processing layer are electrically coupled using through silicon vias (TSV).

20. A head-mounted display (HMD) comprising:

a sensor assembly including:

a sensor layer, of a plurality of stacked layers, including an array of pixels configured to capture one or more first images, and

one or more processing layers of the plurality of stacked layers located beneath the sensor layer, the one or more processing layers configured to:

extract features of one or more objects and their pixel locations from the one or more first images; and

enable a subset of the array of pixels based on the pixel locations to capture one or more second images of the one or more objects;

and

an optical assembly configured to direct image light to an eye box corresponding to a location of a user's eye,

wherein depth information is determined based on the extracted features of the one or more objects.

21. The HMD of claim 20 , wherein the one or more processing layers are configured to:

generate a key frame including information related to the pixel locations of the features of the one or more objects in the one or more first images to a host device coupled with or being part of the HMD;

receive, from the host device, a first map of pixel locations based on the information;

enable the subset of the array of pixels based on the map of pixel locations to capture the one or more second images;

generate, based on the one or more second images, a second map of pixel locations of the features of the one or more objects; and

transmit the second map to the host device,

wherein the second map is transmitted at a higher frame rate than the key frame.

22. The HMD of claim 20 , wherein the subset is a first subset;

wherein the one or more processing layers are further configured to control a second subset of the array of pixels that is not enabled for capturing of the one or more second images to pass pixel values of the one or more first images to the one or more processing layers.

23. The HMD of claim 20 , wherein the sensor assembly further includes an interface connection between each pixel in the array of pixels and logic of at least one processing layer of the one or more processing layers.

24. The HMD of claim 20 , wherein at least one of the one or more processing layers includes a convolutional neural network (CNN) based on an array of memristors for storage of trained network weights.

25. A mobile device comprising:

a sensor assembly including:

a sensor layer, of a plurality of stacked layers, including an array of pixels configured to capture one or more first images, and

one or more processing layers of the plurality of stacked layers located beneath the sensor layer, the one or more processing layers configured to:

extract features of one or more objects and their pixel locations from the one or more first images; and

enable a subset of the array of pixels based on the pixel locations to capture one or more second images of the one or more objects; and

an optical assembly configured to direct image light to an eye box corresponding to a location of a user's eye,

wherein depth information is determined based on the extracted features of the one or more objects.