US 12,462,544 B2
Method and apparatus for image classification
Xiaohui Wang, Guangdong (CN); Guansheng Liang, Guangdong (CN); Qian Sun, Guangdong (CN); Zhijin Zhang, Guangdong (CN); Xing Liu, Jiangsu (CN); Hui Weng, Jiangsu (CN); and Bo Zhou, Jiangsu (CN)
Assigned to Telefonaktiebolaget LM Ericsson (Publ), Stockholm (SE)
Appl. No. 17/777,080
Filed by Telefonaktiebolaget LM Ericsson (publ), Stockholm (SE)
PCT Filed Jan. 23, 2020, PCT No. PCT/CN2020/073965
§ 371(c)(1), (2) Date May 16, 2022,
PCT Pub. No. WO2021/147066, PCT Pub. Date Jul. 29, 2021.
Prior Publication US 2022/0415019 A1, Dec. 29, 2022
Int. Cl. G06V 10/82 (2022.01); G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01)
CPC G06V 10/82 (2022.01) [G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 2201/06 (2022.01)] 16 Claims
OG exemplary drawing
 
1. A method for image classification, the method comprising:
pretraining a first Generative Adversarial Network, GAN, using a plurality of sample images with classification information, the pretraining comprising multiple epochs, each epoch comprising:
training a discriminator of the first GAN with the plurality of sample images with classification information and a plurality of noise images with classification information, while freezing a generator of the first GAN; and
training the generator of the first GAN with random noise and random classification information, while freezing the discriminator of the first GAN, the generator of the first GAN generating the plurality of noise images and the classification information for the plurality of noise images;
the discriminator and the generator of the first GAN being trained iteratively;
a number of iterative times for the discriminator of the first GAN being larger than a number of iterative times for the generator of the first GAN;
receiving an image to be classified;
inputting the image to the discriminator of the first GAN; and
outputting a result indicating real and an index of a predetermined classification, or a result indicating fake.