This project aims to provide practical experience in implementing image colorization using Generative Adversarial Networks (GANs) with a U-Net architecture. Participants will work with a curated dataset of grayscale images paired with their corresponding colored images and will explore the process of training a GAN to predict colors for grayscale inputs.
- Understanding the principles of image colorization.
- Exploring the architecture and capabilities of U-Net GANs.
- Training a U-Net GAN model on a custom dataset for colorization tasks.
- Evaluating the performance of the trained model.
src/: Contains the source code for the project.data.py: Custom data loader for handling the image colorization dataset.ganloss.py: Implementation of GAN loss functions.models.py: Implementation of the U-Net architecture for the generator and discriminator.weight_initializer.py: Script for initializing model weights.
images/: Directory containing the images for training and testing.
You can upload the notebook and the code directly in Colab. Start by exploring the main.ipynb notebook and filling in the TODOs to train and evaluate the image colorization model using the provided dataset. Make sure to install the required dependencies listed in requirements.txt. Happy coding! 🎨🖌️