---
license: mit
language:
- en
- fr
tags:
- computer vision
- deep learning
- pytorch
---
# README for Image Colorization Model

Welcome to the GitHub repository for our Image Colorization model! This model uses deep learning to add vibrant colors to grayscale images, particularly focusing on historical photographs. Below you'll find instructions on how to set up and use this model.

## Requirements
To use this model, you need the following:
- Python 3.8 or later
- PyTorch
- torchvision
- PIL (Pillow)
- scikit-image
- matplotlib
- numpy

You can install the necessary libraries using `pip`:
```bash
pip install torch torchvision Pillow scikit-image matplotlib numpy
```

## Model Architecture
`ColorizationNet` is a neural network model built using PyTorch. It combines the power of ResNet and an upsampling network to colorize grayscale images.

## Setup
First, clone this repository and navigate to the directory:
```bash
git clone [repo-link]
cd [repo-directory]
```

Download the trained model weights (`colorization_md1.pth`) and place them in the repository's root directory.

## Usage

To colorize an image, follow these steps:

1. **Prepare the Image**: Ensure your image is in grayscale format. You can use any standard image format like JPG or PNG.

2. **Run the Model**: Use the following code to colorize your image.

    ```python
    from colorize import ColorizationNet, colorize_single_image
    import torch
    import torch.nn as nn

    # Load the model
    model = ColorizationNet()
    model_path = 'colorization_md1.pth'
    pretrained = torch.load(model_path, map_location=lambda storage, loc: storage)
    model.load_state_dict(pretrained)
    model.eval()  # Set the model to evaluation mode

    # Set the path to your image
    image_path = 'path_to_your_image.jpg'
    save_dir = 'dir_to_save_colorized_image'
    criterion = nn.MSELoss()

    # Colorize the image
    use_gpu = torch.cuda.is_available()
    colorize_single_image(image_path, model, criterion, save_dir, epoch=0, use_gpu=use_gpu)
    ```

3. **View the Results**: The colorized image will be saved in the specified `save_path`. You will find both the original grayscale and the colorized version there.

## Contributions and Issues
Feel free to contribute to this project by submitting pull requests. If you encounter any issues or have questions, please submit them in the issues section of this repository.

## License



Enjoy exploring the vibrant colors of history with our model! 🎨📸