Object detection in optical sensor systems
Object detection techniques for use in conjunction with optical sensors is described. In one or more implementations, a plurality of inputs are received, each of the inputs being received from a respective one of a plurality of optical sensors. Each of the plurality of inputs are classified using machine learning as to whether the inputs are indicative of detection of an object by a respective said optical sensor.
1. A method comprising:
receiving a plurality of inputs, each of the inputs being received from a respective one of a plurality of optical sensors;
classifying each of the plurality of inputs using machine learning as to whether the inputs are indicative of detection of an object by a respective said optical sensor, said classifying including:
using a first classifier to perform the classifying by taking as an input an image that includes both infrared light and ambient light and an image having ambient light subtracted from the infrared light and generating a probability map that describes of detection of the object by respective ones of the plurality of optical sensors; and
using a second classifier to perform the classifying based at least in part on the input having the image that includes both infrared light and ambient light; and
determining a location of the object using a result of the classifying.
2. A method as described in claim 1 , further comprising recognizing a gesture usable to initiate an operation of the computing device based on a result of the determining.
3. A method as described in claim 1 , wherein the input taken by the second said classifier is processed using one or more blob detection techniques.
4. A method as described in claim 1 , wherein the plurality of optical sensors are arranged to form an array.
5. A method as described in claim 4 , wherein the array is part of a sensor-in-pixel functionality of a display device of a computing device.
6. A method as described in claim 1 , wherein the classifying includes use of a discriminative classifier that returns a result as a discrete probability distribution of a set of classes that are indicative of the detection of the object.
7. A method as described in claim 1 , wherein the classifying includes classifying the inputs using a randomized decision forest (RDF) that employs one or more randomized decision trees (RDT).
8. A system comprising:
a plurality of optical sensors;
one or more hardware processors; and
one or more computer readable storage media storing instructions that are executable by the one or more hardware processors to perform operations including:
using a first classifier configured to generate a first probability map that describes a likelihood of detection of an object by respective ones of the plurality of optical sensors, the probability map generated by taking as an input an image that includes both infrared light and ambient light and an image having ambient light subtracted from the infrared light;
using a second classifier configured to generate a second probability map that describes a likelihood of detection of an object by respective ones of the plurality of optical sensors based at least in part on the input having the image that includes both infrared light and ambient light;
determining that an object has been detected using the first and second probability maps; and
determining a location of the object based on the first and second probability maps.
9. A system as described in claim 8 , wherein the second probability map is generated by the second classifier without using the image having ambient light subtracted from the infrared light that is used by the first classifier.
10. A system as described in claim 8 , wherein the using the first classifier and the using the second classifier comprise using machine learning to generate the first probability map and the second probability map, respectively.
11. A system as described in claim 8 , wherein the using the second classifier to generate the second probability map comprises using inputs received from the plurality of optical sensors that have been processed using one or more blob detection techniques.
12. A system as described in claim 8 , wherein the plurality of optical sensors are arranged to form an array.
13. A system as described in claim 12 , wherein the array is part of a sensor-in-pixel functionality of a display device of a computing device.
14. A system as described in claim 8 , wherein the using the first classifier and the using the second classifier comprise using a randomized decision forest (RDF) that employs one or more randomized decision trees (RDT) to generate the first probability map and the second probability map, respectively.
15. One or more computer readable storage media comprising instructions stored thereon that, responsive to execution by a computing device, causes the computing device to perform operations comprising:
generating a first probability map that describes a likelihood of detection of an object by respective ones of a plurality of optical sensors, the probability map generated by taking as an input an image that includes both infrared light and ambient light and an image having ambient light subtracted from the infrared light;
generating a second probability map that describes a likelihood of detection of an object by respective ones of the plurality of optical sensors based at least in part on the input having the image that includes both infrared light and ambient light; and
determining that an object has been detected and a location of the object using the first and second probability maps.
16. One or more computer readable storage media as described in claim 15 , wherein the generating of the first and second probability maps is performed using a first and second randomized decision forest (RDF) that employs one or more randomized decision trees (RDT), respectively.
17. One or more computer readable storage media as described in claim 15 , wherein the plurality of optical sensors are part of a sensor-in-pixel functionality of a display device of the computing device.
18. One or more computer readable storage media as described in claim 15 , wherein the generating the first probability map comprises using a first classifier, and the generating the second probability map comprises using a second classifier.
19. One or more computer readable storage media as described in claim 15 , wherein the generating the second probability map comprises processing the input using one or more blob detection techniques.