Final Project of COMP3065 (HKBU)
This repository hosts code designed for colorization tasks using Convolutional Neural Networks (CNNs) in PyTorch, aiming to convert grayscale images into their vibrant, colorful counterparts. Users have the flexibility to apply the code to their own images or explore its capabilities with the included sample dataset.
Download the dataset zip file from the following link: https://www.kaggle.com/datasets/ashwingupta3012/male-and-female-faces-dataset and unzip it to a location outside the current directory.
Once you have the dataset downloaded, create a folder named dataset and inside this folder, create another folder with the name of your choice. Place all the images in the last created folder.
Run train.py and wait for the training to complete. You can watch a movie or two during this time.
Run inference.py and select the file you desire to colorize (it has to be a human portrait). Result will be displayed in results\result_output.jpeg or jpg