- Final project for the Machine Learning Master's course at University of Groningen
- The focus is on Colourization task using Conditional Generative Adversarial Networks
- Using pre-trained ResNet and UNet
- Some code is borrowed from one of the author's repo
- For running the train script use. Use
--helpto list all the commandline options.
python3 src/train.py --help
- After training, use the following script to compute evaluation metrics on the test set. Use
--helpto list all the commandline options.
python3 src/compute_test_metrics.py --help
- After training, use the following script to colourize the images using the Generator network. Use
--helpto list all the commandline options.
python3 src/generate_results.py --help
- Use notebook src/plot_train_losses.ipynb to visualize training losses
- Use notebook src/visualize_results.ipynb to visualize the generated results from Generator and compare it with the original images.
- The dependencies are available in the requirements.txt