|

PyTorch-Style-Transfer

This repo provides PyTorch Implementation of MSG-Net (ours) and Neural Style (Gatys et al. CVPR 2016), which has been included by ModelDepot. We also provide Torch implementation and MXNet implementation.

Tabe of content

• Real-time Style Transfer using MSG-Net

• Stylize Images using Pre-trained Model

• Train Your Own MSG-Net Model

• Slow Neural Style Transfer

MSG-Net

Multi-style Generative Network for Real-time Transfer [arXiv] [project] Hang Zhang, Kristin Dana @article{zhang2017multistyle,     title={Multi-style Generative Network for Real-time Transfer},     author={Zhang, Hang and Dana, Kristin},     journal={arXiv preprint arXiv:1703.06953},     year={2017} }

Stylize Images Using Pre-trained MSG-Net

1. Download the pre-trained model bash git clone git@github.com:zhanghang1989/PyTorch-Style-Transfer.git cd PyTorch-Style-Transfer/experiments bash models/download_model.sh

2. Camera Demo bash python camera_demo.py demo --model models/21styles.model 

3. Test the model bash python main.py eval --content-image images/content/venice-boat.jpg --style-image images/21styles/candy.jpg --model models/21styles.model --content-size 1024

4. If you don't have a GPU, simply set --cuda=0. For a different style, set --style-image path/to/style. If you would to stylize your own photo, change the --content-image path/to/your/photo. More options:

• --content-image: path to content image you want to stylize.

• --style-image: path to style image (typically covered during the training).

• --model: path to the pre-trained model to be used for stylizing the image.

• --output-image: path for saving the output image.

• --content-size: the content image size to test on.

• --cuda: set it to 1 for running on GPU, 0 for CPU.

Train Your Own MSG-Net Model

1. Download the COCO dataset bash bash dataset/download_dataset.sh

2. Train the model bash python main.py train --epochs 4

3. If you would like to customize styles, set --style-folder path/to/your/styles. More options:

• --style-folder: path to the folder style images.

• --vgg-model-dir: path to folder where the vgg model will be downloaded.

• --save-model-dir: path to folder where trained model will be saved.

• --cuda: set it to 1 for running on GPU, 0 for CPU.

Neural Style

Image Style Transfer Using Convolutional Neural Networks by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge.

python main.py optim --content-image images/content/venice-boat.jpg --style-image images/21styles/candy.jpg

• --content-image: path to content image.

• --style-image: path to style image.

• --output-image: path for saving the output image.

• --content-size: the content image size to test on.

• --style-size: the style image size to test on.

• --cuda: set it to 1 for running on GPU, 0 for CPU.

Acknowledgement

The code benefits from outstanding prior work and their implementations including: - Texture Networks: Feed-forward Synthesis of Textures and Stylized Images by Ulyanov et al. ICML 2016. (code) - Perceptual Losses for Real-Time Style Transfer and Super-Resolution by Johnson et al. ECCV 2016 (code) and its pytorch implementation code by Abhishek. - Image Style Transfer Using Convolutional Neural Networks by Gatys et al. CVPR 2016 and its torch implementation code by Johnson.