proco_ACGAN

Motivation

• How to avoid an afterimage?

Approach #3

• Instead of transfering character itself, what about generating a character and applying user's hand-writting style with Domain transfer?

Structure

• fontto : applying reference for Korean character, plus additional functions needed
• reference : implemented code for acGAN

Limits

• Needs a lot of datasets to train the model and style
• Requires 4~5 hours per user's own style to learn imply the model

Example & Screenshot

Reference

zi2zi