search

sleep3r / Diffusion-Handwriting-Generation.pytorch

Diffusion Model for Online Handwriting Generation
MIT License
1 stars 4 forks source link
diffusion diffusion-models handwritten-text-recognition python pytorch

readme

Diffusion Model for Handwriting Generation [![CodeFactor](https://www.codefactor.io/repository/github/sleep3r/Diffusion-Handwriting-Generation.pytorch/badge)](https://www.codefactor.io/repository/github/sleep3r/Diffusion-Handwriting-Generation.pytorch) [![python](https://img.shields.io/badge/python_3.10-passing-success)](https://github.com/sleep3r/Diffusion-Handwriting-Generation.pytorch/badge.svg?branch=main&event=push)

Data preparation:

Download IAM Handwriting Database and extract it somewhere with the following structure:

DATA_DIR/
├── ascii
├── lineImages
└── lineStrokes

Install:

make install

Train:

First, configure your training in configs/<cfg>.yml:

experiment:
  name: <exp_name>
  work_dir: <work_dir>
  data_dir: <data_dir>

Then, run:

make train CONFIG=<cfg>.yml

References:

Papers
[1] Luhman, Troy, and Eric Luhman. "Diffusion models for handwriting generation." arXiv preprint arXiv:2011.06704 (2020).