Open holmes27 opened 1 year ago
Currently the code is not released but the general idea is to sample Gaussian noises then sent them to the decoders.
@yizhiwang96 Thanks for your prompt reply! I wanted to know how can we give 3-4 characters as input (in svg/ png format) and generate entire font set (svg/ttf) ?
Our method takes ref_nshot
svgs as reference and generates the whole font. ref_nshot
: the number of reference number is fixed for each training. In our experiment, it was 4
. If you want to input your own svgs, you need to first convert it into the format of our dataset, see the instruction in readme.
Thanks a lot! @yizhiwang96 I had a doubt, the sfd files created for each character in the font have StartChar as A (train/test data_set that is available) and while conversion from ttf to sfd the char is set as 'A' in each loop. What is the assumption here?
I have separate svgs for some characters and converted svg to sfd directly for each character
How to run code for random font generation?