sketchy-svg 0.2.6

A png drawing to svg converter

sketch-vectorization

Goal

The goal of this project is to implement the following paper: https://www-sop.inria.fr/reves/Basilic/2016/FLB16/fidelity_simplicity.pdf

We also implemented several new ideas, such as a convolutional neural network with synthetic data augmentation for preprocessing.

Demo

Guess what: the entire library can run in your browser !

If you want a guided tour of how the library works, go here:

If you just want to test it for yourself, with your own drawings, go here:

Install

The library is also available as a pip package:

pip install "sketchy-svg[onnx]"
# or use uv: uv add "sketchy-svg[onnx]"

There is no command line interface. You can easily build your own, to get inspiration look at src/sketchy_svg/viz inside the class Demo

Use locally

To install the dependencies, install uv and run:

uv sync --extra onnx 

To launch the notebooks, run uv run marimo edit ., it should open the notebooks in your browser.

Train the CNN

The CNN denoiser can be trained from notebooks/cnn.py.

1. Install training dependencies

On a machine with a GPU (Linux, default CUDA torch):

uv sync --group train

On a CPU-only machine (e.g. your local machine):

uv sync --group train --extra cpu

2. Run the notebook

uv run marimo edit notebooks/cnn.py

Set USE_RAY = False to train locally, or USE_RAY = True to offload to a remote Ray cluster (set RAY_ADDRESS accordingly). The trained model is exported to src/sketchy_cnn/model.onnx at the end of the notebook.

How the SVG dataset is downloaded

data/svg_dataset.csv is tracked in git so no download is needed in most cases.

If you need to refresh it, the dataset comes from OmniSVG/MMSVG-Icon on HuggingFace. Create a token at https://huggingface.co/settings/tokens and add it to a .env file at the project root:

HF_TOKEN=hf_...

Then use the download button in notebooks/svg_dataset.py.

Documentation

If you want, you can read the Presentation