Train a network to generate handwriting both unconditionally and conditionally based on text input. This code is based on Generating Sequences With Recurrent Neural Networks.
How do I run it?
Run the jupyter notebook in Docker:
- Follow the link in the terminal to run
notebooks/results.ipynbto generate some samples!
Or run it locally:
The results can be run in a jupyter notebook using
Unconditionally Generated Samples
Conditionally Generated Samples
Baruch Tabanpour was here
france will win the world cup
welcome to my house
welcome to canada
Set these env vars to point to your data and model directories:
Then you can train one of the models:
- Unconditional stroke generation:
python unconditional_stroke_model.py ../models/unconditional-stroke-model-1/
- Conditional stroke generation:
python conditional_stroke_model.py ../models/conditional-stroke-model-1/