Skip to content

Latest commit

 

History

History
52 lines (41 loc) · 1.76 KB

File metadata and controls

52 lines (41 loc) · 1.76 KB

Pytorch-RNN-Handwriting-Synthesis

Implementation of a model for handwriting synthesis using Long Short-Term Memory recurrent neural networks in PyTorch. Based on the work of Alex Graves described in this article

Dataset

The dataset used to train this neural network is the IAM On-Line Handwriting Database. In order to train this network you have to register and download the following files:

And the directory structure has to be the following:

    .
    ├── main.py                
    ├── README.md
    ├── parameters.yaml       
    ├── LICENSE
    ├── trained_models          # Trained models
    ├── src                     # Source files
    └── data                    # Data files
        ├── ascii          # text files that contain the written text
        |   ├── a01
        |   |   ├── a01-000
        |   |   |   ├── a01-000u.txt
        |   |   |   └── a01-000x.txt
        |   |   └── ...
        |   └── ...
        └── lineStrokes    # xml files that contain the strokes
            ├── a01
            |   ├── a01-000
            |   |   ├── a01-000u-01.xml
            |   |   └── a01-000u-02.xml
            |   |   └── ...
            |   └── ...
            └── ...
    

Usage

install packages:

pip install -r requirements.txt

See the main script's help text for more information:

python main.py --help