Handwritten Text Generation from Visual Archetypes

This repository contains the reference code and dataset for the paper Handwritten Text Generation from Visual Archetypes.

If you find it useful, please cite it as:

@inproceedings{pippi2023handwritten,
  title={{Handwritten Text Generation from Visual Archetypes}},
  author={Pippi, Vittorio and Cascianelli, Silvia and Cucchiara, Rita},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2023}
}

Installation

conda create --name vatr python=3.9
conda activate vatr
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia
git clone https://github.com/aimagelab/VATr.git && cd VATr
pip install -r requirements.txt

From this folder you have to download the files IAM-32.pickle and resnet_18_pretrained.pth and place them into the files folder.

gdown --folder "https://drive.google.com/drive/u/2/folders/1FGJe2uCuK8T9HrFzY_Zc-KMIo0oPJGGY"

Training

python train.py

Useful arguments:

python train.py
        --feat_model_path PATH  # path to the pretrained resnet 18 checkpoint. If none, the resnet will be trained from scratch
        --dataset DATASET       # dataset to use. Default IAM
        --resume                # resume training from the last checkpoint with the same name
        --wandb                 # use wandb for logging

Pretraining dataset

The model resnet_18_pretrained.pth was pretrained by using this dataset: download link

Generate styled Handwtitten Text Images

To generate all samples for FID evaluation you can use the following script:

python generate_fakes.py --checkpoint files/vatr.pth

To generate a specific text with a given input style folder containing images of handwritten single words you can use the following script:

python generator.py --style-folder "files/style_samples/00" --checkpoint "files/vatr.pth" --output "files/output_00.png" --text "That's one small step for man, one giant leap for mankind ΑαΒβΓγΔδ"

Output for That's one small step for man, one giant leap for mankind ΑαΒβΓγΔδ:

Implementation details

This work is partially based on the code released for Handwriting-Transformers