Awesome

eps-Assign-and-Mix

Official code release for AAAI 2023 paper User-controllable Arbitrary Style Transfer via Entropy Regularization

Model architecture of εAM

Install

git clone https://github.com/cplusx/eps-Assign-and-Mix.git
pip install -r requirements.txt

Dataset

Our model training use MSCOCO for content images and Painter By Number for style images.

Download MSCOCO

mkdir -p Dataset/cocostuff
cd Dataset/cocostuff
wget http://images.cocodataset.org/zips/train2017.zip
wget http://images.cocodataset.org/zips/val2017.zip
unzip train2017.zip
unzip val2017.zip

Download Painters By Number

mkdir -p Dataset/PBN
cd Dataset/PBN
pip install kaggle
kaggle competitions download -c painter-by-numbers -f train.zip
kaggle competitions download -c painter-by-numbers -f test.zip
unzip train.zip
unzip test.zip

Sampling

See epsAM_demo.ipynb

qualitative

Train

python train.py --eps 0