Skip to content

Latest commit

 

History

History
94 lines (68 loc) · 3.66 KB

README.md

File metadata and controls

94 lines (68 loc) · 3.66 KB
Raw

HiGAN+

Introduction

This is a PyTorch implementation of the paper "HiGAN+: Handwriting Imitation GAN with Disentangled Representations" (authored by Ji Gan, Weiwiang Wang*, Jiaxu Leng, Xinbo Gao*. )

HiGAN+ can generate diverse and realistic handwritten text images (with 64-pixel height) conditioned on arbitrary textual contents and calligraphic styles.

Overview of HiGAN+

Overview of HiGAN

Installation & requirements

The current version of the code has been tested with the following environment:

  • Ubuntu 20 or 22
  • Python 3
  • PyTorch 1.11.0

To use the code, download the repository and change into it:

git clone https://github.com/ganji15/HiGANplus.git

cd HiGAN+

You need to applicant the IAM dataset from http://www.fki.inf.unibe.ch/databases/iam-handwriting-database and then extract the handwriting images. For convenience, here we provide the processed h5py files trnvalset_words64_OrgSz.hdf5 testset_words64_OrgSz.hdf5, which should put into the ./data/iam/ directory.

Training & Test

Training HiGAN on the IAM dataset

python train.py --config ./configs/gan_iam.yml

Quantitative Test

python test.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --guided True

  • Main arguments:
    • --config: the configuration file of HiGAN
    • --ckpt: the path of checkpoint, which is stored in the ./runs/ directory after training.
    • --guided: whether to extract styles from reference images. If --guided False, the styles of generated images will be randomly sampled from the standard normal distribution.

Qualitative Evaluation

python eval_demo.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --mode style

  • Main arguments:
    • --config: the configuration file of HiGAN
    • --ckpt: the path of checkpoint, which is stored in the ./runs/ directory after training.
    • --mode: [ rand | style | interp | text ].

Latent-guided synthesis

python eval_demo.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --mode rand Rand

Reference-guided synthesis

python eval_demo.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --mode style Style

Text synthesis

python eval_demo.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --mode text Text

Style interpolation

python eval_demo.py --config ./configs/gan_iam.yml --ckpt ./pretrained/deploy_HiGAN+.pth --mode interp Interp1

On-the-fly plots during training

With this code it is possible to track progress during training with on-the-fly plots. This feature requires Tensorboard, which should be started from the command line:

tensorboard --logdir=./runs

The tensorboard server is now alive and can be accessed at http://localhost:6006.

Some on-the-fly plots are given as the followings: Loss Samples

Citation

If you find our research is helpful, please remember to cite our paper:

@article{gan2022higanplus,
author = {Gan, Ji and Wang, Weiqiang and Leng, Jiaxu and Gao, Xinbo},
title = {HiGAN+: Handwriting Imitation GAN with Disentangled Representations},
year = {2022},
volume = {42},
number = {1},
url = {https://doi.org/10.1145/3550070},
doi = {10.1145/3550070},
journal = {ACM Trans. Graph.}
}

License

HiGAN+ is free for academic research purposes.