	US 12,277,755 B1
	Method, apparatus, device, and medium for image recognition via wireless federated learning
Yanjie Dong, Shenzhen (CN); Luya Wang, Shenzhen (CN); Jia Wang, Shenzhen (CN); Jianqiang Li, Shenzhen (CN); Haijun Zhang, Shenzhen (CN); Fei Yu, Shenzhen (CN); Song Guo, Shenzhen (CN); and Zhongming Liang, Shenzhen (CN)
Assigned to SHENZHEN UNIVERSITY, Shenzhen (CN)
Filed by SHENZHEN UNIVERSITY, Shenzhen (CN)
Filed on Jun. 21, 2024, as Appl. No. 18/750,709.
Int. Cl. G06V 10/778 (2022.01); G06V 10/77 (2022.01); G06V 10/776 (2022.01); G06V 10/94 (2022.01); G06V 10/96 (2022.01); G06V 20/62 (2022.01)

CPC G06V 10/778 (2022.01) [G06V 10/7715 (2022.01); G06V 10/776 (2022.01); G06V 10/94 (2022.01); G06V 10/96 (2022.01); G06V 20/62 (2022.01)]

7 Claims

1. A method for image recognition via wireless federated learning, comprising:

obtaining an image to be recognized and an initial image recognition model;

adjusting a parameter for the initial image recognition model via a preset accelerated mobile federated learning algorithm according to a target momentum factor to obtain a target image recognition model; and

recognizing the image to be recognized through the target image recognition model to obtain text information corresponding to the image to be recognized;

wherein the adjusting the parameter for the initial image recognition model via the preset accelerated mobile federated learning algorithm according to the target momentum factor to obtain the target image recognition model comprises:

obtaining a running dataset of clients, wherein the number of the clients is N, and N is greater than or equal to 1;

performing feature extraction on the running dataset through a preset distributed principal component decomposition algorithm to obtain target running data;

determining target gradient information based on a current model parameter, a link distance, and the target running data;

performing gradient alignment on the initial image recognition model according to current gradient information;

aggregating the initial image recognition model after gradient alignment; and

adjusting the parameter for the initial image recognition model after aggregating via the preset accelerated mobile federated learning algorithm according to the target momentum factor and an iterative auxiliary model parameter to obtain the target image recognition model.