DeOldify - Automagical colorization of black-and-white photos and film

The other day I stumbled across #DeOldify on Twitter, where I saw amazing colorizations of black-and-white photos. All done automagically using deep learning. Since I myself have a bunch of old black-and-white photos of my family that are 100+ years old, I wanted to try this. And many of you did too, so I made this little quick guide.

Right now I know far too little about the project to write about it, but if you want to read more you can find it on GitHub at https://github.com/jantic/DeOldify

How do I get started?

The only thing you really need is that the image you want to colorize is online. In the example I show you cannot upload an image, so you must have a link to it. If you don’t have access to your own web server, you can use services like Imgur or similar. Otherwise it’s basically just click and wait. So let’s go!

Go to https://colab.research.google.com/github/jantic/DeOldify/blob/master/ImageColorizerColab.ipynb and follow the instructions in the video further down in this post. In short it’s about clicking the play button (which appears) inside each [  ] as in the image below until you reach the section where you enter the URL (the link) to your image file. I’ve tested jpg and png and both work. In the video below I’ve shortened the loading times so the clip wouldn’t be too long. Some steps can take a few minutes, so grab a coffee and load up on patience. :) NOTE! If you want to colorize multiple images you don’t need to redo everything in the video for each image; the last step where you enter the image URL and press play is sufficient.

It’s not working!

If you have problems, leave a comment here or on social media and I might be able to help you.

Run on your own hardware?

Of course you can run this on your own hardware instead of Google’s. All the info on how is on GitHub.