Short-term and long-term memory on an edge device
Systems and methods are provided for distributed video storage and search over edge computing devices having a short-term memory and a long-term memory. The method may comprise caching a first portion of data on a first device. The method may further comprise determining, at a second device, whether the first device has the first portion of data. The determining may be based on whether the first piece of data satisfies a specified criterion. The method may further comprise sending the data, or a portion of the data, and/or a representation of the data from the first device to a third device.
1. A dynamic data compression system, comprising:
a group of sensors affixed to a vehicle and operable to detect and capture driving event data, the group of sensors comprising a video data generating sensor;
a controller coupled to the group of sensors and operable to receive one or more data streams indicative of the driving event data from the group of sensors; wherein the controller is operable to:
determine a vehicle operation condition based on the one or more data streams;
determine, based on the vehicle operation condition, whether or not to compress a data stream from the video data generating sensor
determine, based on the vehicle operation condition, that the data stream from the video data generating sensor need not be processed by a neural network; and
process previously recorded image data from the video data generating sensor with the neural network in response to the determination that the data stream from the video data generating sensor need not be processed with the neural network.
2. The dynamic data compression system of claim 1 , wherein the controller is further operable to: process the data stream from the video data generating sensor with a neural network to produce visual inference data, and wherein the determination of the vehicle operation condition is further based on the visual inference data.
3. The dynamic data compression system of claim 1 , wherein the vehicle operation condition comprises an indicator indicating if the vehicle has been stationary for at least a predetermined period of time.
4. The dynamic data compression system of claim 1 , wherein the vehicle operation condition comprises an indicator indicating if a visual scene has not substantially changed for at least a predetermined period of time.
5. A dynamic data compression system, comprising:
a group of sensors affixed to a vehicle and operable to detect and capture driving event data, the group of sensors comprising a video data generating sensor;
a controller coupled to the group of sensors and operable to receive one or more data streams indicative of the driving event data from the group of sensors; wherein the controller is operable to:
determine a vehicle operation condition based on the one or more data streams;
determine, based on the vehicle operation condition, whether or not to compress a data stream from the video data generating sensor;
determine, based on the vehicle operation condition, a priority score for corresponding sensor data, and
determine, based on the priority score being below a processing threshold, that the data stream need not be processed by the neural network.
6. The dynamic data compression system of claim 1 , wherein the previously recorded image data are substantially uncompressed.
7. A dynamic data compression system, comprising:
a group of sensors affixed to a vehicle and operable to detect and capture driving event data, the group of sensors comprising a video data generating sensor;
a controller coupled to the group of sensors and operable to receive one or more data streams indicative of the driving event data from the group of sensors; wherein the controller is operable to:
determine a vehicle operation condition based on the one or more data streams;
pseudo-randomly select packets for compression and transfer to long-term memory;
determine, based on the vehicle operation condition and further based on the pseudo-random selection, whether or not to compress a data stream from the video data generating sensor; and
determine that the data stream from the video data generating sensor need not be processed by a neural network based on the pseudo-random selection.
8. The dynamic data compression system of claim 7 , wherein the controller is further operable to: process previously recorded image data from the video data generating sensor with the neural network in response to the determination that the data stream from the video data generating sensor need not be processed with the neural network.
9. A method comprising:
capturing one or more data streams with a group of sensors affixed to a vehicle, the group of sensors comprising a video data generating sensor;
receiving, by one or more processors within the vehicle, the one or more data streams;
determining, by the one or more processors and based on the one or more data streams, a vehicle operation condition; and
determining, by the one or more processors, and based on the vehicle operation condition, whether or not to compress a data stream from the video data generating sensor;
determining, by the one or more processors and based on the vehicle operation condition, a priority score for corresponding sensor data; and
determining, by the one or more processors and based on the priority score being below a processing threshold, that the data stream from the video data generating sensor need not be processed by a neural network.
10. The method of claim 9 , further comprising:
processing the data stream from the video data generating sensor with a neural network to produce visual inference data, and wherein the determination of the vehicle operation condition is further based on the visual inference data.
11. The method of claim 9 , wherein the vehicle operation condition comprises an indication that the vehicle has been stationary for at least a pre-determined period of time.
12. The method of claim 9 , wherein the vehicle operation condition comprises an indication that that the visual scene has not substantially changed for at least the pre-determined period of time.
13. A method comprising:
capturing one or more data streams with a group of sensors affixed to a vehicle, the group of sensors comprising a video data generating sensor;
receiving, by one or more processors within the vehicle, the one or more data streams;
determining, by the one or more processors and based on the one or more data streams, a vehicle operation condition;
determining, by the one or more processors, and based on the vehicle operation condition, whether or not to compress a data stream from the video data generating sensor;
determining, by the one or more processors and based on the vehicle operation condition, that the data stream from the video data generating sensor need not be processed by a neural network; and
processing previously recorded image data from the video data generating sensor with the neural network in response to the determination that the data stream from the video data generating sensor need not be processed with the neural network.
14. The method of claim 13 , wherein the previously recorded image data are substantially uncompressed.
15. A method comprising:
capturing one or more data streams with a group of sensors affixed to a vehicle, the group of sensors comprising a video data generating sensor;
receiving, by one or more processors within the vehicle, the one or more data streams;
determining, by the one or more processors and based on the one or more data streams, a vehicle operation condition;
pseudo-randomly selecting, by the one or more processors, packets for compression and transfer to long-term memory;
determining, by the one or more processors, and based on the vehicle operation condition and further based on the pseudo-random selection, whether or not to compress a data stream from the video data generating sensor; and
determining, by the one or more processors that the data stream from the video data generating sensor need not be processed by a neural network based on the pseudo-random selection.
16. The method of claim 15 , further comprising:
processing previously recorded image data from the video data generating sensor with the neural network in response to the determination that the data stream from the video data generating sensor need not be processed with the neural network.