© 2019 Fourtoes

ColouriseSG – Machine learning.

ColouriseSG is a project by the Data Science and Artificial Intelligence Division, GovTech Singapore.

Andrew Tan, Quantitative Analyst
Preston Lim, Software Engineer
Tan Kai Wei, Data Scientist

From the website: Our deep learning model performs best on higher resolution images that prominently feature human subjects and natural scenery. We do not store images that you upload for colourisation.

Its so simple to use just upload an image and in a few seconds a colour image is produced. I was amazed. Granted you can argue about subject accuracy, but remember this is machine learning, artificial intelligence teaching itself from a data set. Yes, teaching itself! You can read more about that here.

It does come with a disclaimer which is fair enough…

Note: The purpose of colourisation is to generate an image with colours that are plausible. It by no means guarantees that the colourised image is an accurate representation of the actual snapshot in time.

Here are some of the examples I put through it. These are the more successful ones with the most colour attributed to the subject matter. Some crackers for sure.

And heres a modern medium format film image from last years ECW by the talented Guillaume d’Hubert. This one worked so well.

I have to thank my friend and fellow photographer Patrick Ballanger for pointing out this amazing web tool yesterday. Be quick as theres talk that this is only online for a short while. Lets hope they make it into an App.

Accurate? Maybe not.
Fun? Definitely.

Makes you look at images in a totally new way.


One Comment