DPCT

DPCT: Disentangling Personal Style and Content for Chinese Handwriting Generation with Transformer

Machine Vision and Learning Lab, Computer Science and Information Engineering, National Chung Cheng University
國立中正大學資訊工程學系機器視覺學習實驗室

It's my master's thesis and implementation code.

Introduction

Based on the implementation of SDT, the training process is modified, and the content encoder is used to extract potential content features of style samples for other subsequent comparative learning applications.

Thanks for their great work !!

Get Started

Git Clone

git clone https://github.com/AUDOSt0ck1ng/DPCT.git

Create Conda Env

conda create --name DPCT python=3.8
y
conda activate DPCT

Install dependencies

cd ./DPCT
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
y
pip3 install -r requirements.txt

Prepare Data, Use the Offered Data from SDT:

ln -s "your_decompression_path" "./data"