Neural Style Transfer with PyTorch

Description

Pytorch implementation of the Neural Style Transfer algorithm. Link to the original paper by Gatys et al.

How to run

First, install dependencies (a new python virtual environment is recommended).

# clone project   
git clone https://github.com/visualCalculus/neural-style-transfer

# install project   
cd neural-style-transfer
pip install -e .
pip install -r requirements.txt

Next, navigate to nst folder and run train.py with appropriate content and style image paths.

# module folder
cd nst

CONTENT_DIR={path to content image}
STYLE_DIR={path to style image}
OUTPUT_DIR={path to save result image}

# run module
python train.py --use_gpu=true --content_dir=${CONTENT_DIR} \
--style_dir=${STYLE_DIR} --input_image="content" \
--output_dir=${OUTPUT_DIR} --iterations=100 \ 
--alpha=1 --beta=1000000 \ 
--style_layer_weight=1.0

Results

Citation

@misc{visualCalculus2021nst,
  title={neural-style-transfer},
  author={Tejomay, Abhiroop},
  publisher={GitHub}
  journal={GitHub repository},
  year={2021},
  howpublished={\url{https://github.com/visualCalculus/neural-style-transfer}}
}