US 12,354,327 B2
Apparatus and methods for generating edge ground truth data for a federated system architecture using machine learning processes
Thomas Svantesson, Linköping (SE)
Assigned to QUALCOMM Incorporated, San Diego, CA (US)
Filed by Arriver Software AB, Linköping (SE)
Filed on Sep. 16, 2022, as Appl. No. 17/946,854.
Prior Publication US 2024/0096058 A1, Mar. 21, 2024
Int. Cl. G06V 10/764 (2022.01); G06V 10/26 (2022.01)
CPC G06V 10/764 (2022.01) [G06V 10/26 (2022.01)] 28 Claims
OG exemplary drawing
 
1. An apparatus comprising:
a communications interface;
a non-transitory, machine-readable storage medium storing instructions; and
at least one processor coupled to the communications interface, and the non-transitory, machine-readable storage medium, the at least one processor configured to execute the instructions to:
obtain, from a database, (i) a first dataset of a first trained machine learning process associated with a first type of object detection and recognition and (ii) a second dataset of a second trained machine learning process associated with a second type of object detection and recognition, the first dataset including at least a first value of a parameter of the first trained machine learning process;
transmit, to a first edge device and one or more other edge devices of a plurality of edge devices via the communications interface, a message including the first dataset and the second dataset, the message causing the first edge device and the one or more other edge devices to:
apply the first trained machine learning process to respective sensor data in accordance with the first dataset, the respective sensor data being associated with each edge device of the first edge device and the one or more other edge devices;
apply the second trained machine learning process to the respective sensor data in accordance with the second dataset;
perform operations that retrain the first trained machine learning process based in part on first respective ground truth data generated based on the application of the first trained machine learning process to the respective sensor data and second respective ground truth data generated based on the application of the second trained machine learning process to the respective sensor data, the retrained first trained machine learning process being associated with a second value of the parameter of the first trained machine learning process; and
generate a respective dataset including the second value of the parameter;
receive, from the first edge device and the one or more other edge devices, the respective dataset including the second value of the parameter; and
update the first dataset with the respective dataset received from the first edge device and the one or more other edge devices, an updated portion of the first dataset including the second value of the parameter.