US 12,307,751 B2
Training method of neural network model and associated device
Chia-Chun Cheng, HsinChu (TW)
Assigned to Realtek Semiconductor Corp., HsinChu (TW)
Filed by Realtek Semiconductor Corp., HsinChu (TW)
Filed on Jun. 30, 2022, as Appl. No. 17/853,914.
Claims priority of provisional application 63/217,769, filed on Jul. 2, 2021.
Prior Publication US 2023/0005246 A1, Jan. 5, 2023
Int. Cl. G06V 10/776 (2022.01); G06T 7/70 (2017.01); G06V 10/82 (2022.01); G06V 40/10 (2022.01)
CPC G06V 10/776 (2022.01) [G06T 7/70 (2017.01); G06V 10/82 (2022.01); G06V 40/10 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30196 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A training method of a neural network model, comprising:
receiving image data, wherein the image data comprises a plurality of frames, and for a plurality of first frames in the plurality of frames, the image data further comprises detection data of each first frame, the detection data comprises position of at least one person within the corresponding first frame, and the detection data does not comprise any serial number of the person; and for a plurality of second frames in the plurality of frames, the image data further comprises person search data of each second frame, and the person search data comprises position and serial number of at least one person within the corresponding second frame;
using the neural network model to perform a person recognition operation on the plurality of frames to generate a recognition result; and
using a plurality of loss functions to process the recognition result of each frame, the detection data of each first frame and the person search data of each second frame, for adjusting parameters of the neural network model.