Systems and methods for domain-specific enhancement of real-time models through edge-based learning
Techniques described herein provide for the use of a heterogeneous artificial intelligence/machine learning (“AI/ML”) architecture, in which relatively complex AI/ML, techniques may be used in conjunction with more lightweight AI/ML techniques in order to leverage the accuracy of relatively complex AI/ML techniques with the reduced processing power and/or time requirements of more lightweight AI/ML techniques. A teacher model system may utilize processing resource and/or time-intensive AI/ML techniques and/or models in order to determine classifications associated with source data, and may provide such classifications to a student model system that may utilize the classifications in accordance with less processing resource and/or time-intensive AI/ML techniques in order to accurately classify sensor data in real time or near-real time.
1 . A system, comprising:
a far cloud device configured to:
receive first sensor data associated with a first device that is associated with a particular geographical region;
utilize a first set of operations to identify a first set of classifications associated with the first sensor data;
generate or modify one or more models associated with the particular geographical region based on the first set of classifications;
select, from a plurality of Multi-Access/Mobile Edge Computing (“MEC”) devices, a particular MEC device that is associated with the particular geographical region; and
provide the one or more models to the selected particular MEC device;
wherein the particular MEC device is configured to:
receive second sensor data associated with a second device that is associated with the particular geographical region;
utilize the one or more models generated based on the first set of classifications and further utilize a second set of operations, that is different from the first set of operations, to identify a second set of classifications associated with the second sensor data; and
provide the second set of classifications to an autonomous or semi-autonomous vehicle, wherein operation of the autonomous or semi-autonomous vehicle is based on the second set of classifications.
2 . The system of claim 1 , wherein the first sensor data and the second sensor data include video data, wherein the first set of operations includes a tracking operation performed on the video data of the first sensor data, and wherein the second set of operations does not include the tracking operation performed on the video data of the second sensor data.
3 . The system of claim 2 , wherein the tracking operation includes identifying a particular object depicted in the video data and a path associated with the object.
4 . The system of claim 1 , wherein the first set of operations include one or more post-processing operations performed on the first sensor data, wherein the second set of operations do not include the one or more post-processing operations performed on the second sensor data.
5 . The system of claim 1 , wherein the first set of operations include one or more filtering operations performed on the first sensor data, wherein the second set of operations do not include the one or more filtering operations performed on the second sensor data.
6 . The system of claim 1 , wherein the first and second sensor data include a same type of data.
7 . A non-transitory computer-readable medium, storing a plurality of processor-executable instructions to:
receive, by a far cloud device, first sensor data associated with a first device that is associated with a particular geographical region;
utilize, by the far cloud device, a first set of operations to identify a first set of classifications associated with the first sensor data;
generate or modify, by the far cloud device, one or more models associated with the particular geographical region based on the first set of classifications;
select, by the far cloud device and from a plurality of Multi-Access/Mobile Edge Computing (“MEC”) devices, a particular MEC device that is associated with the particular geographical region;
provide, by the far cloud device, the one or more models to the selected particular MEC device;
receive, by the MEC device, second sensor data associated with a second device that is associated with the particular geographical region;
utilize, by the MEC device, the one or more models generated based on the first set of classifications and further utilize a second set of operations, that is different from the first set of operations, to identify a second set of classifications associated with the second sensor data; and
provide, by the MEC device, the second set of classifications to an autonomous or semi-autonomous vehicle, wherein operation of the autonomous or semi-autonomous vehicle is based on the second set of classifications.
8 . The non-transitory computer-readable medium of claim 7 , wherein the first and second sensor data include a same type of data.
9 . The non-transitory computer-readable medium of claim 8 , wherein the first and second sensor data each include at least one of:
video data,
still image data, or
audio data.
10 . The non-transitory computer-readable medium of claim 8 , wherein the first sensor data and the second sensor data include video data, wherein the first set of operations includes a tracking operation performed on the video data of the first sensor data, and wherein the second set of operations does not include the tracking operation performed on the video data of the second sensor data.
11 . The non-transitory computer-readable medium of claim 10 , wherein the tracking operation includes identifying a particular object depicted in the video data and a path associated with the object.
12 . The non-transitory computer-readable medium of claim 7 , wherein the first set of operations include one or more post-processing operations performed on the first sensor data, wherein the second set of operations do not include the one or more post-processing operations performed on the second sensor data.
13 . The non-transitory computer-readable medium of claim 7 , wherein the first set of operations include one or more filtering operations performed on the first sensor data, wherein the second set of operations do not include the one or more filtering operations performed on the second sensor data.
14 . A method, comprising:
receiving, by a far cloud device, first sensor data associated with a first device that is associated with a particular geographical region;
utilizing, by the far cloud device, a first set of operations to identify a first set of classifications associated with the first sensor data;
generating or modifying, by the far cloud device, one or more models associated with the particular geographical region based on the first set of classifications;
selecting, by the far cloud device and from a plurality of Multi-Access/Mobile Edge Computing (“MEC”) devices, a particular MEC device that is associated with the particular geographical region;
providing, by the far cloud device, the one or more models to the selected particular MEC device;
receiving, by the MEC device, second sensor data associated with a second device that is associated with the particular geographical region;
utilizing, by the MEC device, the one or more models generated based on the first set of classifications and further utilize a second set of operations, that is different from the first set of operations, to identify a second set of classifications associated with the second sensor data; and
providing, by the MEC device, the second set of classifications to an autonomous or semi-autonomous vehicle, wherein operation of the autonomous or semi-autonomous vehicle is based on the second set of classifications.
15 . The method of claim 14 , wherein the first sensor data and the second sensor data include video data, wherein the first set of operations includes a tracking operation performed on the video data of the first sensor data, and wherein the second set of operations does not include the tracking operation performed on the video data of the second sensor data.
16 . The method of claim 15 , wherein the tracking operation includes identifying a particular object depicted in the video data and a path associated with the object.
17 . The method of claim 14 , wherein the first set of operations include one or more post-processing operations performed on the first sensor data, wherein the second set of operations do not include the one or more post-processing operations performed on the second sensor data.
18 . The method of claim 14 , wherein the first set of operations include one or more filtering operations performed on the first sensor data, wherein the second set of operations do not include the one or more filtering operations performed on the second sensor data.
19 . The method of claim 14 , wherein the first and second sensor data include a same type of data.
20 . The method of claim 19 , wherein the first and second sensor data each include at least one of:
video data,
still image data, or
audio data.