makiminar/image-colorization

Image colorization using CNN, AE and VAE.

Image colorization

The goal is to create and implement models that take grayscale images as input and predict colourized images as output. The colourization will focus on art images - paintings, portraits, etc. The framework used for this semestral work is Pytorch.

Some subgoals need to be completed, specifically:

• create an appropriate art dataset,
• desing and implement ML models,
• evaluate the results on testing data.

Dataset

I downloaded several separate art datasets and then concatenated them into one of 6883 images. The first dataset I downloaded is from \url{https://www.kaggle.com/datasets/ikarus777/best-artworks-of-all-time}. The author downloaded paintings from the top 50 most influentual artists from artchallenge.ru website. The second dataset I downloaded was from \url{https://data.mendeley.com/datasets/289kxpnp57/1}, which is a collection of of portraits.

How to run the notebook

You can either download the notebook and run it in your environment, or you can see the work in this Google Colab: https://colab.research.google.com/drive/1h6a5Aq2k6MoWAQwJmQmm2ZDA7C76j9WD?pli=1&authuser=5#scrollTo=C1hIENcSQ1Tj. (only for members of the shared disc)