This project's aim is to colorize grayscale images through the help of convolutional autoencoders. Primarily, we are concerned with colorizing indoor scenes. The images used for training/val/test are from the MIT Indoor Scenes dataset.
The data can be downloaded from Kaggle using this link.
The script for the project is usable with some changes to file directory on your system.
Pre-trained model weights from this project are located here.