🖼️ Photo Colorization with DDColor

This repository demonstrates how to colorize black-and-white photos using DDColor, a deep learning model for high-quality automatic image colorization.
It includes a simple Jupyter Notebook to test the model on sample grayscale images.

---

📘 Overview

• Framework: ModelScope (with DDColor)
• Interface: Jupyter Notebook (tested in VS Code 1.105)
• OS Tested: Windows 10
• Date Tested: November 2025
• Purpose: Simple demonstration of automatic colorization using the DDColor model through ModelScope.

---

🧩 Features

• Automatically colorize black-and-white photos.
• Uses the pre-trained DDColor model from ModelScope.
• Simple notebook-based workflow for experimentation.
• No manual model downloads — ModelScope handles model loading automatically.

---

⚙️ Installation

1. Clone this repository

   git clone https://github.com/cyamahat/Photo-Colorization-DDColor.git
   cd Photo-Colorization-DDColor

2. Create and activate a virtual environment

   python -m venv venv
   venv\Scripts\activate   # On Windows

3. Install dependencies

   pip install "modelscope[cv]" timm opencv-python matplotlib

   (This includes the necessary packages for this notebook: ModelScope, DDColor, OpenCV, and Matplotlib.)

4. Launch Jupyter Notebook

   jupyter notebook

   Then open the notebook file Photo-Colorization-DDColor.ipynb.

---

🧪 Usage

1. Open the notebook in VS Code or Jupyter Lab.
2. Run the cells step-by-step to:
   • Import the required packages.
   • Load the DDColor model from ModelScope.
   • Select a sample black-and-white image.
   • Generate and display the colorized output.

---

🖼️ Example

Input (B/W) → Output (Colorized)

Black & White Photograph	Colorized

Sample image: Bardeen, Shockley, and Brattain (1948)
(Used for demonstration under Wikipedia Commons licensing.)

---

Other examples: Churchill (1941) | Chaplin (Modern Times) | Battle of Iwo Jima (1945) | Solvay conference (1927) | Françoise Hardy (New York, 1965) |Farm house (Carroll County, Georgia, 1941) | Sailor in Breech (1908), Bardeen, Shockley, and Brattain (1948) | Einstein (1951)

---

🧠 Model Reference

DDColor: Deep Colorization with Detail Decoupling

• Paper: https://arxiv.org/abs/2212.11613
• Official Model: https://modelscope.cn/models/damo/cv_ddcolor_image-colorization
• Original Repo: https://github.com/piddnad/DDColor

---

💡 Notes

• Tested with Visual Studio Code 1.105, Python 3.12, and Jupyter on Windows 10.
• GPU acceleration is optional — CPU works fine for small images (slower).
• ModelScope automatically downloads the DDColor model on first use.

---

🪪 License

This project follows the same license as DDColor.
Sample image © Wikipedia / Wikimedia Commons.

---

🤝 Acknowledgements

• DDColor implementation by piddnad/DDColor
• Model provided by ModelScope
• Sample image from Wikipedia: John Bardeen, William Shockley and Walter Brattain at Bell Labs, 1948