	US 11,875,584 B2
	Method for training a font generation model, method for establishing a font library, and device
Jiaming Liu, Beijing (CN); and Licheng Tang, Beijing (CN)
Assigned to Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Filed by Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Filed on Feb. 28, 2022, as Appl. No. 17/682,131.
Claims priority of application No. 202111056558.8 (CN), filed on Sep. 9, 2021.
Prior Publication US 2023/0114293 A1, Apr. 13, 2023
Int. Cl. G06V 30/244 (2022.01); G06V 30/19 (2022.01); G06T 11/20 (2006.01)

CPC G06V 30/245 (2022.01) [G06T 11/203 (2013.01); G06V 30/1916 (2022.01); G06V 30/19127 (2022.01); G06V 30/19147 (2022.01); G06V 30/19173 (2022.01)]

16 Claims

1. A method for training a font generation model, comprising:

inputting a source-domain sample character into the font generation model to obtain a first target-domain generated character, wherein the font generation model is a cyclic network generation model and comprises a first generation model and a second generation model;

inputting the first target-domain generated character into a font recognition model to obtain a target adversarial loss of the font generation model;

updating a model parameter of the first generation model for multiple rounds according to the target adversarial loss until the first generation model is determined to satisfy a model stability condition, wherein the model stability condition comprises that a current number of updates of the first generation model reaches a set number of times; and

inputting the first target-domain generated character into a pre-trained character classification model to obtain a character loss of the font generation model; inputting the first target-domain generated character and the target-domain sample character into the character classification model to obtain a feature loss of the font generation model; and updating the model parameter of the first generation model according to the character loss and the feature loss.