US 11,734,955 B2
Disentangled representation learning generative adversarial network for pose-invariant face recognition
Xiaoming Liu, Okemos, MI (US); Luan Quoc Tran, Haslett, MI (US); and Xi Yin, East Lansing, MI (US)
Assigned to BOARD OF TRUSTEES OF MICHIGAN STATE UNIVERSITY, East Lansing, MI (US)
Appl. No. 16/648,202
Filed by BOARD OF TRUSTEES OF MICHIGAN STATE UNIVERSITY, East Lansing, MI (US)
PCT Filed Sep. 18, 2018, PCT No. PCT/US2018/051552
§ 371(c)(1), (2) Date Mar. 17, 2020,
PCT Pub. No. WO2019/056000, PCT Pub. Date Mar. 21, 2019.
Claims priority of provisional application 62/560,001, filed on Sep. 18, 2017.
Prior Publication US 2020/0265219 A1, Aug. 20, 2020
Int. Cl. G06V 40/16 (2022.01); G06N 20/00 (2019.01); G06F 18/214 (2023.01); G06V 10/24 (2022.01)
CPC G06V 40/172 (2022.01) [G06F 18/214 (2023.01); G06N 20/00 (2019.01); G06V 10/242 (2022.01); G06V 40/165 (2022.01); G06V 40/168 (2022.01)] 28 Claims
OG exemplary drawing
 
1. A method for identifying a subject using imaging, the method comprising:
receiving an image depicting a subject to be identified;
applying a trained Generative Adversarial Network (GAN) to the image to generate an identity representation of the subject, wherein the GAN comprises a discriminator and a generator, the generator including at least one encoder and a decoder;
wherein generating the identity representation of the subject includes weighing features associated with each of a plurality of images for the subject using learning coefficients, the identity representation being disentangled from pose variations;
generating at least one synthetic image based on the identity representation, the at least one synthetic image including a rotated pose of the subject, the rotated pose of the subject in the at least one synthetic image being different from a pose of the subject in the image;
identifying the subject using the identity representation; and
generating a report indicative of the subject identified.