US 12,444,185 B2
Facial recognition based on converted spiking neural network
Yu Qian, Irmo, SC (US); and Youzhi Tang, Columbia, SC (US)
Assigned to UNIVERSITY OF SOUTH CAROLINA, Columbia, SC (US)
Filed by UNIVERSITY OF SOUTH CAROLINA, Columbia, SC (US)
Filed on May 8, 2023, as Appl. No. 18/313,563.
Claims priority of provisional application 63/438,548, filed on Jan. 12, 2023.
Claims priority of provisional application 63/346,088, filed on May 26, 2022.
Prior Publication US 2023/0386195 A1, Nov. 30, 2023
Int. Cl. G06V 10/82 (2022.01); G06V 10/77 (2022.01); G06V 10/774 (2022.01); G06V 10/94 (2022.01); G06V 40/16 (2022.01)
CPC G06V 10/82 (2022.01) [G06V 10/774 (2022.01); G06V 10/94 (2022.01); G06V 40/172 (2022.01); G06V 10/7715 (2022.01)] 36 Claims
OG exemplary drawing
 
1. Method for facial recognition for automatically identifying a person from facial images, comprising:
training an Artificial Neural Network (ANN)-based facial recognition machine learning model, based on inputs of a preexisting large-scale facial image dataset for facial recognition, to learn the association between different facial images of the same person;
normalizing the parameters of the ANN-based facial recognition machine learning model using a randomly selected subset of facial images which is smaller than the preexisting large-scale dataset used to train the ANN-based facial recognition machine learning model;
converting the trained ANN-based facial recognition machine learning model with normalized parameters into a trained Spiking Neural Network (SNN)-based facial recognition model; and
operating the trained SNN-based facial recognition machine learning model to process further input data images thereto, to determine and output the identification of the person corresponding with the input images.