Animating Doodles with Autoencoders

This repo contains the code for my blog post on creating animations of doodles using autoencoders trained on synthetic data.

Requirements

modules.ipynb and autoenc_models.ipynb contain the code for building the pytorch models.
lines_dataset.ipynb contains the code for making the synthetic dataset.
Run train.ipynb after setting a save directory name to train a CoordConv autoencoder with a latent space size of 64. You can change this and try out other models as well.
Play around with a trained model to make gifs using the analysis.ipynb notebook.

A sample grid of images for the tree gif is also available as tree.bmp. Use this to create some gifs using the autoencoder you trained.

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
.gitignore		.gitignore
README.md		README.md
analysis.ipynb		analysis.ipynb
autoenc_models.ipynb		autoenc_models.ipynb
autoenc_models.py		autoenc_models.py
lines_curves_dataset.py		lines_curves_dataset.py
lines_dataset.ipynb		lines_dataset.ipynb
modules.ipynb		modules.ipynb
modules.py		modules.py
train.ipynb		train.ipynb
tree.bmp		tree.bmp