| --- |
| title: imageup |
| emoji: 🎨 |
| colorFrom: blue |
| colorTo: green |
| sdk: docker |
| app_port: 7860 |
| pinned: false |
| --- |
| |
| # 🎨 ImageUP - AI Image Colorization |
|
|
|  |
|  |
|  |
|  |
|  |
|
|
| ## 📖 Project Overview |
|
|
| **ImageUP** is a full-stack web application that leverages Deep Learning to automatically colorize black-and-white images. |
|
|
| The project integrates a **Flask** backend with **OpenCV's** DNN module to run Caffe models for image processing. The frontend is designed with **HTML5** and **Tailwind CSS**, providing a modern, drag-and-drop interface. The entire application is containerized using **Docker** for consistent deployment. |
|
|
| --- |
|
|
| ## 📂 Repository Structure |
|
|
| | File / Directory | Description | |
| | :--- | :--- | |
| | `app.py` | **Main Application**: Flask server handling routes, image processing, and inference logic. | |
| | `download_models.py` | **Setup Script**: Python script to automatically download the required Caffe model files. | |
| | `Dockerfile` | **Container Config**: Instructions to build the Docker image with Python 3.11 and system dependencies. | |
| | `requirements.txt` | **Dependencies**: List of Python libraries (`flask`, `opencv-python`, `numpy`, `gunicorn`). | |
| | `templates/index.html` | **Frontend**: User interface built with HTML and styled using Tailwind CSS (via CDN). | |
| | `models/` | **Model Storage**: Directory where `colorization_release_v2.caffemodel` and related files are stored. | |
| | `static/` | **Assets**: Contains `uploads/` for input images and `results/` for processed outputs. | |
|
|
| --- |
|
|
| ## 🛠️ Tech Stack & Dependencies |
|
|
| The project is built using the following technologies: |
|
|
| | Category | Technology | Usage | |
| | :--- | :--- | :--- | |
| | **Language** | Python 3.11 | Core programming language. | |
| | **Web Framework** | Flask | Handles HTTP requests and routing. | |
| | **Computer Vision** | OpenCV (`cv2`) | Image manipulation and Deep Learning inference. | |
| | **Data Processing** | NumPy | Matrix operations for image tensors. | |
| | **Frontend** | HTML5, JavaScript | Structure and client-side logic. | |
| | **Styling** | Tailwind CSS | Utility-first CSS framework for the UI. | |
| | **Infrastructure** | Docker | Containerization and deployment. | |
| | **Server** | Gunicorn | WSGI HTTP Server for production. | |
|
|
| --- |
|
|
| ## ✨ Key Features |
|
|
| * **AI-Powered Colorization**: Utilizes the Zhang et al. (2016) Caffe model for realistic color prediction. |
| * **Drag & Drop Interface**: User-friendly upload zone with immediate visual feedback. |
| * **Real-time Processing**: Fast inference pipeline using OpenCV's optimized DNN backend. |
| * **Comparison View**: Side-by-side display of the original grayscale and the new colorized image. |
| * **Automatic Model Fetching**: `download_models.py` handles the retrieval of large model weights. |
|
|
| --- |
|
|
| ## 🚀 Usage Instructions |
|
|
| ### Prerequisites |
| * **Docker** (Recommended) |
| * *OR* **Python 3.x** with `pip` installed. |
|
|
| ### Option 1: Run with Docker (Easiest) |
| 1. **Build the Image** |
| ```bash |
| docker build -t imageup-app . |
| ``` |
| 2. **Run the Container** |
| ```bash |
| docker run -p 5000:5000 imageup-app |
| ``` |
| Access the application at `http://localhost:5000`. |
| |
| ### Option 2: Run Locally (Python) |
| 1. **Install Dependencies** |
| ```bash |
| pip install -r requirements.txt |
| ``` |
| 2. **Download Models** |
| The model files are too large for git. Run this script to fetch them: |
| ```bash |
| python download_models.py |
| ``` |
| 3. **Start the Server** |
| ```bash |
| python app.py |
| ``` |
| Open your browser and navigate to `http://localhost:5000`. |
| |
| --- |
|
|
| ## 🧠 Model Details |
|
|
| The application uses the **Colorful Image Colorization** algorithm. |
| * **Input**: Lightness channel (L) from the LAB color space. |
| * **Output**: Predicted 'a' and 'b' color channels. |
| * **Process**: The system combines the original L channel with the predicted ab channels to reconstruct the final color image. |
|
|
| --- |
|
|
| ## 👤 Author |
|
|
| **Rishabh** |
| * [GitHub Profile](https://github.com/rixabhh) |
| * [LinkedIn](https://www.linkedin.com/in/rishabbh/) |
|
|
