Segmentation-guided Layer-wise Vectorization with Gradient Fills

This repo requires PyTorch and torchvision to work. Please refer to Start Locally | PyTorch on how to install PyTorch.

We have tested on PyTorch 2.1.1 with Python 3.10.13. The environment is saved to env.yml.

This work is largely inspired by LIVE.

Install DiffVG

The directory DiffVG contains a forked version of the original DiffVG. We edited DiffVG/pydiffvg/save_svg.py to save radial gradient parameters.

With the current working directory changed to DiffVG:

conda install -y numpy scikit-image 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

Install SGLIVE Dependencies

pip install opencv-python==4.5.4.60 
pip install torchmetrics
pip install easydict

Run

With the current working directory changed to SGLIVE:

python main.py --config config/sglive.yaml --experiment experiment_16x1 --signature noto_u1f61a --target data/noto_u1f61a.png