S2WAT - GitLite

KO | EN

gitlite — search

#go #proxy #github #markdown #redis #claude #animation #react-native #anthropic #awesome-list

S2WAT

(★ 44)

Official PyTorch implementation of “S2WAT: Image Style Transfer via Hierarchical Vision Transformer using Strips Window Attention”

File Explorer

ZIP Download

- __init__.py
- configuration.py
- s2wat.py
- transformer_components.py
- transformer_tools.py
- fig1_title.png
- fig2_architecture.png
- fig3_st_examples.png
- fig4_content_leak.png
- fig5_st_examples_box.png
- fig6_st_examples.png
data_preprocess.py
dataset_sampler.py
emvironment.yml
net.py
README.md
scheduler.py
test.py
tools.py
train.py

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

Command Glossary

Commands referenced in this DOCs, explained below.

🔍

`conda env`

View Details ▼

Manage conda environments.

conda env create {{[-f|--file]}} {{path/to/file}}

Create an environment from an environment file (YAML, TXT, etc.):

conda env remove {{[-n|--name]}} {{environment_name}}

Delete an environment and everything in it:

conda env update {{[-f|--file]}} {{path/to/file}} --prune

Update an environment based on an environment file:

🔍

`conda`

View Details ▼

Package, dependency, and environment management for any programming language.
Some subcommands such as `create` have their own usage documentation.
See also: `mamba`.

conda create {{[-n|--name]}} {{environment_name}} {{python=3.9 matplotlib}}

Create a new environment, installing named packages into it:

conda info {{[-e|--envs]}}

List all environments:

conda activate {{environment_name}}

Activate an environment:

🔍

`python3`

View Details ▼

This command is an alias of `python`.

tldr python

View documentation for the original command:

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<center class='half'>
    <img src='./pics/fig1_title.png' alt='fig1_title.png' title='fig1_title.png' width='' />
</center>

Official [PyTorch](https://pytorch.org) implementation of the paper:

[**S2WAT: Image Style Transfer via Hierarchical Vision Transformer using Strips Window Attention**](https://arxiv.org/abs/2210.12381)

[Chiyu Zhang](https://alienzhang1996.github.io/), Xiaogang Xu*, [Lei Wang](https://orcid.org/0000-0002-6301-0949), [Zaiyan Dai](https://orcid.org/0000-0003-0838-6318),  [Jun Yang\*](https://dblp.org/pid/181/2799-25.html)

S2WAT is a universal style transfer method consisting of three parts: a) an encoder based on a hierarchical vision Transformer, b) a transfer module based on a Transformer decoder, and c) a decoder based on a mirrored VGG. Introducing window-based hierarchical vision Transformers, such as Swin Transformer, to image style transfer directly will result in *locality problem*. S2WAT overcomes this problem with Strips Window Attention (SpW Attention) and is capable of  alleviating the *content leak* issue.

## Architecture

<center class='half'>
    <img src='./pics/fig2_architecture.png' alt='fig2_architecture.png' title='fig2_architecture.png' width='' />
</center>

## Style Transfer Examples

<center class='half'>
    <img src='./pics/fig5_st_examples_box.png' alt='fig5_st_examples_box.png' title='fig3_st_examples.png' width='' />
</center>

<center class='half'>
    <img src='./pics/fig6_st_examples.png' alt='fig6_st_examples.png' title='fig3_st_examples.png' width='' />
</center>

## Results under Content Leak

<center class='half'>
    <img src='./pics/fig4_content_leak.png' alt='fig4_content_leak.png' title='fig4_content_leak.png' width='' />
</center>

## Dependencies

- Python = 3.8.13
- PyTorch = 1.10.0
- CUDA = 11.3
- cuDNN = 8.2.0

Optionally, if you are a `conda` user, you can execute the following command in the root directory of this repository to create a new environment with all dependencies installed.

```python
conda env create -f environment.yaml
```

## Pretrained Models & Datasets

### Pretrained Models

If you want to use the pre-trained model to perform style transfer or train a new model taking the pre-trained parameters as the initial, please download the pre-trained files from Google Drive as follows:

- <a id="VGG19" href="https://drive.google.com/file/d/1nJt6nnEIjBfQMzbH9__TrLJfmHqkaHjy/view?usp=share_link">VGG19</a>
- <a id="Pre-trained" href="https://drive.google.com/file/d/16Ihs_J9ULYSze2lL5cmptvMyy-ZYJ9kN/view?usp=share_link">Pre-trained model</a>

Put VGG19 to the ```pre_trained_models``` directory and the pre-trained model to the ```pre_trained_models/checkpoint``` directory under the root of this repository. Then execute the following commands ([Training ](#Training) or [Testing](#Testing)) in the root of the repository.

### Datasets

The datasets of content images and style images are as follows:

|                  Content Dataset                  |                     Style Dataset                      |
| :-----------------------------------------------: | :----------------------------------------------------: |
| [MS_COCO 2014](https://cocodataset.org/#download) | [WikiArt](https://www.kaggle.com/c/painter-by-numbers) |

We select 80000 pictures from COCO and WikiArt respectively to build content and style datasets. Then preprocess the content and style datasets before training with ```data_preprocess.py``` in which images are resized to 512 on the shorter side and then randomly cropped to 224 x 224. You can use the [preprocessed version](https://www.kaggle.com/datasets/alienzhang/styletransfer224) of the datasets directly or execute the following commands to build one.

```python
python3 data_preprocess.py --source_dir ./source --target_dir ./target
```

- ```source_dir```: Directory to the images to be processed
  - ```source_dir``` needs a subfolder of any name to load images, such as ```./source/images```
- ```target_dir```: Directory to save the processed images

## Training<a id="Training"> </a>

To train S2WAT by yourself, you can run the following commands. Before running the commands, please make sure:

- the pre-trained [VGG19](#VGG19), *vgg_normalised.pth*, should be placed to ```vgg_dir```.
- the [pre-trained model](#Pre-trained), *checkpoint_40000_epoch.pkl* as default, should be placed to ```checkpoint_import_path```, if you want to train a model taking a checkpoint as the initial parameter.

```python
# Train from scratch
python3 train.py --content_dir ./input/Train/Content --style_dir ./input/Train/Style --vgg_dir ./pre_trained_models/vgg_normalised.pth --epoch 40000 --checkpoint_save_interval 10000 --loss_count_interval 400

# Train from a checkpoint
python3 train.py --content_dir ./input/Train/Content --style_dir ./input/Train/Style --vgg_dir ./pre_trained_models/vgg_normalised.pth --epoch 40000 --checkpoint_save_interval 10000 --loss_count_interval 400 --resume_train True --checkpoint_import_path ./pre_trained_models/checkpoint/checkpoint_40000_epoch.pkl
```

- ```content_dir```: Directory path to a batch of content images
- ```style_dir```: Directory path to a batch of style images
- ```vgg_dir```: Directory path to the pre-trained VGG19
- ```epoch```: The number of iterations of the training has
- ```checkpoint_save_interval```: Directory path to save a checkpoint
- ```loss_count_interval```: The interval to calculate average losses
- ```resume_train```: Whether or not to train with a checkpoint as the initial parameters
- ```checkpoint_import_path```: Directory path to the importing checkpoint

## Testing<a id="Testing"> </a>

To test the style transfer performance of the pre-trained model. Before running the following command, please make sure:

- the [pre-trained model](#Pre-trained), *checkpoint_40000_epoch.pkl*, are placed to the ```checkpoint_import_path```.
- the ```input_dir``` contains two subfolders, *Content* and *Style*, and each subfolder has loaded images.

```python
python3 test.py --input_dir ./input/Test --output_dir ./output --checkpoint_import_path ./pre_trained_models/checkpoint/checkpoint_40000_epoch.pkl
```

- ```input_dir```: Directory path to a batch of content and style images which are loaded in *Content* / *Style* subfolders respectively
- ```output_dir```: Directory to save the output images
- ```checkpoint_import_path```: Directory path to the importing checkpoint

## Contact

If you have any questions, please contact alienzhang19961005@gmail.com.