Videoconferencing using hybrid edge/cloud inference with machine-learned systems

1. A computer-implemented method comprising:

obtaining, by one or more computing devices via one or more sensors, sensor data representative of an environment including one or more users;

selecting, by the one or more computing devices using a machine-learned system comprising one or more machine-learned models, at least a portion of the sensor data that is associated with the one or more users;

generating, by the one or more computing devices using the machine-learned system, an intermediate representation of the selected portion of the sensor data that is associated with one or more users, wherein generating the intermediate representation of the selected portion of the sensor data is at least one step in a multi-step identification process;

transmitting, by the one or more computing devices to a remote computing system, the intermediate representation, a request for a user identification based on the intermediate representation, and a header including process data identifying a next step in the multi-step identification process, the remote computing system configured to continue performing additional steps of the multi-step identification process beginning at the identified next step using the intermediate representation and the machine-learned system; and

receiving, from the remote computing system, a user identifier for the one or more users.

2. The computer-implemented method of claim 1 , wherein generating the intermediate representation further comprises:

selecting, based on one or more criteria, a step in the multi-step identification process; and

performing steps in the multi-step identification process until the multi-step identification process reaches an end of the selected step and produces an intermediate representation of the selected portion of the sensor data associated with the selected step, wherein a plurality of steps in the multi-step identification process each produce a distinct intermediate representation of the selected portion of the sensor data.

3. The computer-implemented method of claim 2 , wherein the one or more criteria include a bandwidth currently available to transmit data from the one or more computing devices to the remote computing system.

4. The computer-implemented method of claim 2 , wherein the one or more criteria include a computational resource level available at the remote computing system.

5. The computer-implemented method of claim 1 , wherein the at least one step in the multi-step identification process is performed using a first machine-learned model of the machine-learned system, and the additional steps in the multi-step identification process are performed by a second machine-learned model of the machine-learned system.

6. The computer-implemented method of claim 1 , wherein the user identifier includes a name of a user.

7. The computer-implemented method of claim 6 , further comprising:

displaying, in a display associated with the one or more computing devices, the name of a user in the one or more users.

8. The computer-implemented method of claim 1 , wherein the intermediate representation includes one or more machine-learned primitives associated with the multi-step identification process.

9. The computer-implemented method of claim 8 , wherein the one or more machine-learned primitives include one or more object localization computations.

10. The computer-implemented method of claim 8 , wherein the one or more machine-learned primitives include one or more partial computations for the multi-step identification process.

11. The computer-implemented method of claim 1 , wherein transmitting the intermediate representation and the request for the user identification comprises:

transmitting local metadata associated with the image data to the remote computing system.

12. The computer-implemented method of claim 1 , wherein the method further comprises, prior to transmitting the intermediate representation and the request for the user identification, verifying one or more security and privacy criteria.

13. The computer-implemented method of claim 1 , wherein the sensor data is uncompressed image data.

14. The computer-implemented method of claim 13 , wherein the uncompressed image data is captured during a video conference.

15. The computer-implemented method of claim 14 , wherein the user identifier is received while the video conference is ongoing.

16. A device comprising:

a processor;

one or more sensors;

a display; and

a memory that stores instructions that, when executed by the processor, cause the device to perform operations, the operations comprising:

obtaining, via the one or more sensors, image data of an environment including one or more users;

selecting, using one or more machine-learned models, at least a portion of the sensor data that is associated with one or more users;

generating, using the one or more machine-learned models, an intermediate representation of the selected portion of the sensor data that is associated with the one or more users, wherein generating the intermediate representation of the selected portion of the sensor data is at least one step in a multi-step identification process;

transmitting, to a remote computing system, the intermediate representation and a request for a user identification based on the intermediate representation, a request for a user identification based on the intermediate representation, and a header including process data identifying a next step in the multi-step identification process, the remote computing system configured to continue performing additional steps of the multi-step identification process beginning at the identified next step using the intermediate representation and the one or more machine-learned models; and

receiving, from the remote computing system, a user identifier for the one or more users.

17. The device of claim 16 , wherein the user identifier includes a name of a user and the operations further comprise:

displaying, in the display, the name for a user in the one or more users.

18. A non-transitory computer-readable medium storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising:

obtaining, via one or more sensors, sensor data of a scene including one or more users;

generating, using the one or more machine-learned models, an intermediate representation of the sensor data, wherein generating the intermediate representation of the sensor data is at least one step in a multi-step identification process;

transmitting, to a remote computing system, the intermediate representation and a request for a user identification based on the intermediate representation, and a header including process data identifying a next step in the multi-step identification process, the remote computing system configured to continue performing additional steps of the multi-step identification process beginning at the identified next step using the intermediate representation and one or more machine-learned models; and

receiving, from the remote computing system, a facial recognition identifier for the one or more users.

19. The non-transitory computer-readable medium of claim 18 , wherein the facial recognition identifier includes a name of a user and the operations further comprise:

displaying, in a display associated with the one or more computing devices, the name for a user in the one or more users.