Skeleton Autoencoder

This repo contains an implementation of the algorithm described in the paper, "Using Autoencoders to Generate Skeleton-Based Typography" presented at EvoMUSART 2023.

Project website: CDV Website

Installation

This project has been tested with Ubuntu 18.04. If you're installing on another operating system, you may encounter issues.

1. The first step is to install the diffvg project. We used a simplified version of the original project in order to remove the issues regarding the Tensorflow, as we are not using it in this project.

   git clone https://github.com/lmagoncalo/diffvg
   cd diffvg
   git submodule update --init --recursive
   conda install -y pytorch torchvision -c pytorch
   conda install -y numpy
   conda install -y scikit-image
   conda install -y -c anaconda cmake
   conda install -y -c conda-forge ffmpeg
   pip install svgwrite
   pip install svgpathtools
   pip install cssutils
   pip install numba
   pip install torch-tools
   pip install visdom
   python setup.py install
   

2. Then extract the dataset.zip to extract our custom dataset into your environment.
3. Run the main.py file to train the models.
4. Finally, run the latent_exploration.py file to create a similar image to the following one to visualise the learned latent space of the trained Autoencoder.

Issues

If you encounter any issues please forward them to lgoncalo(at)dei.uc.pt