search

kaonashi-tyc / zi2zi

Learning Chinese Character style with conditional GAN
https://kaonashi-tyc.github.io/2017/04/06/zi2zi.html
Apache License 2.0
2.52k stars 470 forks source link

Can i use a new model to get the same effect, such as VAE #50

Open SHUSHENGQIGUI opened 6 years ago

SHUSHENGQIGUI commented 6 years ago

The result of experiment is surprising ,so i have an question: can i use an alternative way to get the same effect, such as VAE. because my recent research is VAEs, and VAE is also a generative model., I wanna try it . could you give me any advice? Thank you!!

ouguozhen commented 6 years ago

@SHUSHENGQIGUI could you tell me what's your input of the experiment?

SHUSHENGQIGUI commented 6 years ago

@ouguozhen i think my input is a lot of characters which have one specified font, the labels are the same characters which have different font . the model try to make the output(reconstruction) look like the label.That is , i made the transfer of font .Then ,i can input a character into the model ,the output is the character, whose font is like the label of training model.

ouguozhen commented 6 years ago

@SHUSHENGQIGUI Thank you for your reply! however, is it not that each input image can only contain two characters?

Instantnoodles-madman commented 4 years ago

The result of experiment is surprising ,so i have an question: can i use an alternative way to get the same effect, such as VAE. because my recent research is VAEs, and VAE is also a generative model., I wanna try it . could you give me any advice? Thank you!!

can you offer me the Pretained model ? i can't download it 1459984437@qq.com