Colourization of Grayscale images using UNet

How to setup the project?

Project Organization

├── README.md          <- The top-level README for developers using this project.
├── data
│   └── raw            <- Folder containg the original coloured images.
│
├── experiments        <- Trained and serialized models saved during the training after 5 epochs.
│
├── predictions        <- A video showing improvements in colorization over 30 epochs. 
|
├── references         <- Contains a readme file containg all the refrences.
│
├── reports            <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures        <- Figures and Plots.
│
├── requirements.txt   <- The requirements file for reproducing the analysis environment.
│
├── unet               <- Architecture of UNET.
|   ├── unet_model.py  <- UNET model
|   └── unet_parts.py  <- Parts of UNET
|
├── dataloader.py      <- Dataloader for loading train and validation data.
|
├── inference.py       <- Takes path of model and grayscale image and generates a colour image from it.
|
├── train.py           <- Train, validate and save the model.
|
└── saving.py          <- Saving the fig, model checkpoints. 

---

1. Install the requirements using pip install -r requirement.txt.
2. To train the model and validate the model:
   • Run: python train.py
3. To test the model on particular grayscale image on a particular model checkpoint:
   • Run: python inference.py --p_img "path_to_image" --p_model "path_to_model"