AI Image Captioner

⭐ Star us on GitHub — it motivates us a lot!

An intelligent image captioning tool using Azure OpenAI's GPT-4V model, specifically designed for training diffusion models. This tool preprocesses images and generates high-quality captions optimized for AI training.

🎯 About

AI Image Captioner is a Python library designed to provide comprehensive support for image captioning using Azure OpenAI's GPT-4V model. It adheres to high standards of:

• Modularity: Different components can function independently
• Testability: Improved separation of concerns
• Maintainability: Clear structure and organization
• Reliability: Robust error handling and retry mechanisms

✨ Features

Image Preprocessing

• Automatic resizing to 1024x1024
• Aspect ratio preservation with padding
• RGB conversion (alpha channel removal)
• Standardized JPEG output
• Sequential naming (image0.jpg, image1.jpg, etc.)

Caption Generation

• Intelligent captioning using GPT-4V
• Gender and attribute detection
• Consistent formatting
• No introductory phrases
• Focus on main subjects
• Comma-separated elements

Batch Processing

• Multiple file format support
• Progress tracking
• Error handling
• Retry mechanism for API failures

Output Options

• Individual caption files (.txt)
• Dataset JSON file
• CSV export
• Customizable prefix/suffix

📦 Prerequisites

• Python 3.8+
• Azure OpenAI API access with GPT-4V deployment
• Required Python packages (see requirements.txt)

🚀 Installation

1. Clone the repository:

git clone https://github.com/CroissanStudioDev/ai-image-captioner.git
cd ai-image-captioner

2. Create and activate a virtual environment:

# Create virtual environment
python -m venv .venv

# Activate on Linux/macOS
source .venv/bin/activate

# Activate on Windows
.venv\Scripts\activate

3. Install the package in development mode:

pip install -e .

4. Create a .env file with your Azure OpenAI credentials:

AZURE_OPENAI_API_KEY=your_api_key_here
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com
AZURE_OPENAI_DEPLOYMENT_NAME=your_deployment_name

💡 Usage

Basic Usage

Process images with default settings:

python -m ai_image_captioner.cli --input-folder input --output-folder output

Advanced Usage

Add style prefix and quality suffix:

python -m ai_image_captioner.cli \
    --input-folder input \
    --output-folder output \
    --prefix "in the style of TOK" \
    --suffix "high quality"

Output Format

output/
├── image0.jpg          # Preprocessed image
├── image0.txt          # Caption file
├── image1.jpg
├── image1.txt
├── ...
├── captions.csv        # All captions in CSV format
└── dataset.json        # Complete dataset information

Caption Examples

woman with pink hair wearing orange sunglasses, gray jumpsuit with orange accents
man in blue shirt and jeans standing with arms crossed
young girl holding a red balloon, wearing a white dress

🛠️ Development

Project Structure

ai-image-captioner/
├── src/
│   └── ai_image_captioner/
│       ├── __init__.py
│       ├── cli.py
│       ├── image_captioner.py
│       └── preprocessor.py
├── tests/
├── .env.example
├── .gitignore
├── LICENSE
├── README.md
├── requirements.txt
└── setup.py

Code Style

The project follows PEP 8 guidelines and uses Black for formatting:

# Format code
black src/ai_image_captioner

# Check style
flake8 src/ai_image_captioner

❗ Troubleshooting

Common Issues

1. ModuleNotFoundError: Ensure proper installation:

pip install -e .

2. Import Error: Verify virtual environment:

which python  # Should point to .venv/bin/python

3. API Errors: Check .env configuration

🤝 Contributing

Contributions are welcome! Please read CONTRIBUTING.md for details on our code of conduct and the process for submitting pull requests.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

💬 Support

For support:

• Open an issue in the GitHub repository
• Contact: serge@croissanstudio.ru

---

Made with ❤️ by Croissan Studio