traindataset

[pengyaru]

Hi, @ecnuycxie (1)Can you send me the train dataset and val dataset?
(2)Does train dataset has 400 classes once in one training process? if not, what is the minimum number of classes used for training? (3)And what's the length of one class style font in train dataset for example 14 with only "苟利国家生死以岂因祸福避趋之"? (4)And can you tell me the 990 commonly used Chinese characters (content)? (5)Can you send me the pretrained model, .pth file?

[ecnuycxie]

Hi, @pengyaru, I will answer these questions one by one. (1) Due to license restrictions, I cannot send you my dataset (more than 400 .ttf files). But you can download .ttf from 方正字库 which provides thousands of .ttf files. (2) Yes, we have trained 400 styles of DGFont at the same time. However, you can also use a small number of styles (for example, 100 or less). (3) For each style, we use 990 characters (800 for training and 190 for testing). (4) You can obtain these 990 commonly used Chinese characters from http://www.aige.co.uk/resources.asp?id=1. (5) We are sorry that we cannot provide our pre-trained model yet.

[haooxia]

This link is invalid, can you provide it again.

https://www.aige.co.uk/resources.asp?id=1