Stable Diffusion Kanji Generator

A comprehensive project for generating kanji images using stable diffusion models, with data preparation, model training, and deployment using Hugging Face.

Table of Contents

• Directory Structure
• Installation
• Data Preprocessing
• Training the Model
• Generating Images
• Model Theory
• Usage
• Contributing
• License

Directory Structure

project-root/
├── data/
│   ├── dataset/
│   │   └── dataset128.json
│   ├── models/
│   │   └── checkpoints/
│   ├── processed/
│   │   ├── definitions/
│   │   ├── images128/
│   │   └── svg128/
│   └── raw/
│       ├── kanjidic2/
│       └── kanjivg/
├── scripts/
│   ├── generate_image.py
│   ├── preprocess_data.py
│   └── train_model.py
├── src/
│   ├── data_preprocessing/
│   │   ├── __init__.py
│   │   ├── dataset_builder.py
│   │   ├── decompress_data.py
│   │   ├── kanjidic_parser.py
│   │   └── svg_to_pixel.py
│   └── model/
│       ├── cache/
│       │   ├── locks/
│       │   ├── models--CompVis--stable-diffusion-v1-4/
│       │   ├── models--openai--clip-vit-large-patch14/
│       │   ├── models--runwayml--stable-diffusion-v1-5/
│       │   └── models--stabilityai--stable-diffusion-2-1-base/
│       ├── __init__.py
│       └── training.py
├── pyproject.toml
└── README.md

Installation

This project uses Poetry for dependency management.

1. Clone the repository

   git clone https://github.com/username/kanji-stable-diffusion.git
   cd kanji-stable-diffusion

2. Install Poetry (if not already installed)

   curl -sSL https://install.python-poetry.org | python3 -

3. Install project dependencies

   poetry install

Data Preprocessing

Preprocess the kanji data using the provided script:

poetry run python scripts/preprocess_data.py

This script will handle the parsing of KanjiDic2 data, SVG processing, and dataset building.

Training the Model

Train the stable diffusion model with the following command:

poetry run python scripts/train_model.py

This will use the preprocessed data and train the model, saving checkpoints in the data/models/checkpoints/ directory.

Generating Images

Generate kanji images using the trained model:

poetry run python scripts/generate_image.py

Model Theory

This project utilizes stable diffusion models to generate high-quality kanji images. Stable diffusion is a deep learning method that generates images by gradually denoising random Gaussian noise, making it particularly effective for generating detailed and coherent images like kanji characters.

Usage

After training, use the model to generate kanji images:

from src.model.training import generate_image

prompt = "A kanji character for 'mountain' in cursive style"
image = generate_image(prompt)
image.save("kanji_mountain_cursive.png")

Contributing

Contributions are welcome! Please fork the repository and submit a pull request with your improvements.

License

This project is licensed under the MIT License.