We release the code used in the following paper:

Progressive Painterly Image Harmonization from Low-level Styles to High-level Styles [arXiv]

Li Niu, Yan Hong, Junyan Cao, Liqing Zhang

Accepted by AAAI 2024

Our method can harmonize a composite image from low-level styles to high-level styles. The results harmonized to the highest style level have sufficiently stylized foregrounds, but also take the risk of content distortion and artifacts. The users can select the result harmonized to the proper style level.

• Linux
• Python 3.9
• PyTorch 1.10
• NVIDIA GPU + CUDA

• Clone this repo:

git clone https://github.com/bcmi/ProPIH-Painterly-Image-Harmonization.git

• Prepare the datasets as in PHDNet.

• Install PyTorch and dependencies:

conda create -n ProPIH python=3.9
conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch -c conda-forge

• Install python requirements:

pip install -r requirements.txt

• Download pre-trained VGG19 from Baidu Cloud (access code: pc9y) or Dropbox. Put it in ./<checkpoints_dir>/pretrained

• Train ProPIH:

Modify the content_dir and style_dir to the corresponding path of each dataset in train.sh.

cd scripts
bash train.sh

The trained model would be saved in ./<checkpoints_dir>/<name>/. If you want to load a model and resume training, add --continue_train and set the --epoch XX in train.sh. It would load the model ./<checkpoints_dir>/<name>/<epoch>_net_G.pth. For example, if the model is saved in ./AA/BB/latest_net_G.pth, the checkpoints_dir should be ../AA/, the name should be BB, and the epoch should be latest.

• Test ProPIH:

Our pre-trained model is available in Baidu Cloud (access code: azir) or Dropbox. Put it in ./<checkpoints_dir>/pretrained. We provide some test examples in ./examples.

cd scripts
bash test.sh

The output results would be saved in ./output. Some results are shown below. We can see that from stage 1 to stage 4, the composite images are harmonized progressively from low-level styles (color, simple texture) to high-level styles (complex texture).

• Awesome-Image-Harmonization
• Awesome-Image-Composition