Distributed computing resource management
Systems and methods for distributed event detection. Sensor data is synchronized, and an on-board detector included in a vehicle detects an event from the synchronized sensor data. The synchronized sensor data is transmitted to a remote system. At the remote system, a remote detector detects the event and generates a remote label from at least a data subset of the synchronized sensor data.
1. A method for distributed event detection, comprising:
at a vehicle, sampling synchronized interior and exterior image data;
with an on-board detector executing on-board the vehicle, the on-board detector comprising a first computational model type:
detecting a first event based on the synchronized interior and exterior image data;
in response to first event detection, selecting a data subset from the synchronized interior and exterior image data based on the first event; and
in response to first event detection, transmitting the data subset to a remote system; and
with a remote detector executing at the remote system, determining a remote label based on the data subset, wherein the remote detector is associated with the first event, comprises a second computational model type different from the first computational model type, and is more accurate than the on-board detector.
2. The method of claim 1 , wherein the remote detector is configured to detect the first event.
3. The method of claim 1 , further comprising:
at a training module executing at the remote system, training an instance of the onboard detector based on the remote label and the data subset; and
updating the on-board detector executing at the vehicle based on the trained instance of the on-board detector.
4. The method of claim 3 , wherein updating the on-board detector comprises: providing weights of the trained instance of the on-board detector to the onboard detector, wherein the on-board detector is updated to include the weights received from the remote system.
5. The method of claim 1 , wherein the data subset is transmitted from the vehicle to the remote system in response to detection of the first event.
6. The method of claim 1 , wherein the remote detector and the on-board detector are different types of detectors, and the remote detector is more accurate than the on-board detector.
7. The method of claim 1 , wherein the vehicle includes a plurality of on-board detectors, and the method further comprising: selecting at least one on-board detector to disable based on operational data of the vehicle, wherein the operational data indicates at least one of temperature, processing load, available storage capacity, and power usage of the vehicle computing subsystem.
8. The method of claim 1 , wherein the on-board detector and the remote detector are configured to detect vehicle collisions.
9. The method of claim 1 , wherein the on-board detector and the remote detector are configured to generate labels for first-notice-of-loss (FNOL) generation.
10. The method of claim 1 , wherein the on-board detector and the remote detector are configured to generate labels for driver behavior.
11. The method of claim 1 , further comprising:
in response to first event detection: generating an on-board label for the first event with the on-board detector; and
transmitting the on-board label to the remote system in association with the data subset.
12. The method of claim 11 , further comprising, in response to generation of the onboard label:
selecting additional vehicle sensor data to send to the remote system based on the on-board label; and
transmitting the selected additional vehicle sensor data to the remote system in association with the data subset;
wherein the remote detector determines the remote label based on the data subset and at least one of the on-board label and the additional vehicle sensor data.
13. A vehicle computing system comprising:
an interior camera configured to generate interior image data of a vehicle;
an exterior camera configured to generate exterior image data of the vehicle;
a sensor synchronization subsystem configured to generate synchronized image data of the interior camera and the exterior camera;
an on-board detector comprising a first computational model type, and configured to detect a first event from the synchronized image data; and
a communication system configured to transmit the synchronized image data to a remote system;
wherein the remote system comprises a remote detector configured to detect the first event and to generate a remote label from at least a data subset of the synchronized image data;
wherein the communication system is configured to receive weights of a trained instance of the on-board detector from the remote system and update the on-board detector to include the received weights; and
wherein the remote system comprises a training module that is configured to train the instance of the on-board detector based on the remote label and the synchronized image data.
14. The system of claim 13 , wherein the communication system is configured to transmit the synchronized image data to the remote system in response to detection of the first event.
15. The system of claim 13 , wherein the remote detector and the on-board detector are different types of detectors, wherein the remote detector is more accurate than the on-board detector.
16. A vehicle computing system comprising:
an interior camera configured to generate interior image data of a vehicle;
an exterior camera configured to generate exterior image data of the vehicle;
a sensor synchronization subsystem configured to generate synchronized image data of the interior camera and the exterior camera;
an on-board detector comprising a first computational model type, and configured to detect a first event from the synchronized image data; and
a communication system configured to transmit the synchronized image data to a remote system;
wherein the remote system comprises a remote detector configured to detect the first event and to generate a remote label from at least a data subset of the synchronized image data; and
wherein at least one of the communication system and the on-board detector is configured to select a portion of the synchronized image data send to the remote system based on the detected first event, and to control transmission of the selected portion to the remote system.
17. The system of claim 16 , wherein the communication system is configured to transmit the synchronized image data to the remote system in response to detection of the first event.
18. The system of claim 16 , wherein the remote detector and the on-board detector are different types of detectors, wherein the remote detector is more accurate than the on-board detector.
19. A system comprising:
an on-board detector of a vehicle configured to detect a first event and to generate an associated on-board label from synchronized interior image data and exterior image data of the vehicle; and
a remote detector of a remote system, the remote detector configured to detect the first event and to generate a remote label from at least a portion of the synchronized interior image data and exterior image data of the vehicle, wherein the remote detector receives the interior image data and the exterior image data from the vehicle;
wherein the remote detector and the on-board detector are different types of detectors;
wherein the remote detector is more accurate than the on-board detector; and
wherein in response to generation of the on-board label, the on-board detector is configured to select additional vehicle sensor data to send to the remote system based on the on-board label, and transmit the selected additional vehicle sensor data to the remote system in association with the interior image data and the exterior image data.
20. A system comprising:
an on-board detector of a vehicle configured to detect a first event and to generate an associated on-board label from synchronized interior image data and exterior image data of the vehicle; and
a remote detector of a remote system, the remote detector configured to detect the first event and to generate a remote label from at least a portion of the synchronized interior image data and exterior image data of the vehicle;
wherein the remote detector receives the interior image data and the exterior image data from the vehicle, wherein the remote detector and the on-board detector are different types of detectors;
wherein the remote detector is more accurate than the on-board detector; and
wherein the on-board detector is configured to update weights used by the onboard detector with updated weights received from the remote system, the updated weights being weights of a trained instance of the on-board detector trained by the remote system.