adaptive-style-transfer: A Python repository from chendicao

A Style-Aware Content Loss for Real-time HD Style Transfer

Artsiom Sanakoyeu*, Dmytro Kotovenko*, Sabine Lang, Björn Ommer
Heidelberg University
In ECCV 2018 (Oral)

Please click on image for HiRes results.

Website: https://compvis.github.io/adaptive-style-transfer
Paper: https://arxiv.org/abs/1807.10201

Requirements:

python 2.7
tensorflow 1.2.
PIL, numpy, scipy, os
tqdm, argparse

Inference routine:

Simplest van Gogh example.

To launch the inference on van Gogh style first download the pretrained model named 'model_van-gogh' here and sample photographs from here. Extract the model to ./models/ folder and sample photographs to ./data/ folder. Finally run the following code:

CUDA_VISIBLE_DEVICES=0 python main.py \
                 --model_name=model_van-gogh \
                 --phase=inference \
                 --image_size=1280

Stylized photographs are stored in the folder ./models/model_van-gogh/inference_ckpt300000_sz1280/

Additional settings.

If you want to run the code on your own data please specify additional parameter --ii_dir defining a path to folder containing your target images. You can specify multiple folders separated with commas (don't use spaces!).

You can also change the resolution of the image you generate by changing parameter --image_size, please specify

To save generated images in custom folder add parameter --save_dir.

If you don't have a GPU big enough for current model you can set parameter CUDA_VISIBLE_DEVICES="" to use CPU.

All together it looks like this:

CUDA_VISIBLE_DEVICES="" python main.py \
                 --model_name=model_van-gogh \
                 --phase=inference \
                 --image_size=1280 \
                 --ii_dir ../my_photographs1/,../my_photographs2/ \
                 --save_dir=../save_processed_images_here/

Additional artists.

We have pretrained models for the following artists: Paul Cezanne, El-Greco, Paul Gauguin, Wassily Kandinsky (Василий Кандинский), Ernst Ludwig Kirchner, Claude Monet, Berthe Morisot, Edvard Munch, Samuel Peploe, Pablo Picasso, Jackson Pollock, Nicholal Roerich (Николай Рерих), Vincent van Gogh. A few more artists will be added in the future.

You can download the models (link) and extract them to the folder ./models/. Now use the name of the folder as the model_name parameter, for instance for Picasso execute:

CUDA_VISIBLE_DEVICES=0 python main.py \
                 --model_name=model_picasso \
                 --phase=inference \
                 --image_size=1280

Training.

To start training you need a content dataset with photographs and style dataset with images representing artistic style. We have trained our models on Places2 dataset, in particular we've used the Places365-Standard high-res train dataset images(105GB). Please specify path to it using the parameter --ptcd which stands for path_to_content_dataset.

The dataset representing a particular artistic period of Vincent van Gogh (it was automatically collected using "Road with Cypress and Star, 1890" painting as query) can be downloaded from here, extract corresponding folder to the folder ./data/.

Now we can launch the training process:

CUDA_VISIBLE_DEVICES=1 python main.py \
                 --model_name=model_van-gogh_new \
                 --batch_size=1 \
                 --phase=train \
                 --image_size=768 \
                 --lr=0.0002 \
                 --dsr=0.8 \
                 --ptcd=/path/to/Places2/data_large \
                 --ptad=./data/vincent-van-gogh_road-with-cypresses-1890/

We also provide style images we've used to train Claude Monet model: monet_water-lilies-1914.tar.gz.

Video.

Coming soon.

chendicao/adaptive-style-transfer