Convert Black & White Images to Color
Hello World,
As you know AI is my passion, as so why not to develop AI for the masses? today I want to interduce you to deep learning that can colorize images.
First, I need to say a big thank you to for his amazing blog about AI and lots of interesting stuff that help me develop my AI skills.
Some Wiki information
I won’t get into much details about how Convolutional Neural Networks work, you can find all of the information at http://richzhang.github.io/colorization/
The idea is to use Deep Learning to learn from lots of images. the process is achieved by converting them from RGB to Lab Color Space.
Later using the L (light intensity) channel that is the same for RGB and BW and tries and find the other A&B color channels (A&B channels are the colors in Lab Color Space)
Combine all to one RGB and voila! you have colored image.
If its too complicated just move to the demo that I have else go to the link provided.
So that about a demo?
Now for the fun stuff.
I created a website by using Azure Web App, Azure Function. Python and OpenCV.
A new website was born where you can convert any Black & White image you have into color.
good use for the technology is old photos that you have from 1900 of your family, this is the reason I started the project from the first place as I have tons on old BW images.
As for the website. it’s a simple ASP.Net that I created on my free time so it’s not a bug-free.
The only limitations I made to the website is its only gets JPG and only less than 2MB in size.
Look at the example below on how deep learning is doing an amazing job, not 100% right but I think that a human will have the same difficulty or probably the same accuracy level in some points.
Original Photo
Converted it to BW
Result by AI
Now Go and try it yourself!
Related Posts
About The Author
Lior
Cloud Monitoring specialist, with more than 10 years of experience in providing brode wide monitoring solutions for SMP and Enterprises. I work at CloudValley, Microsoft's partner of the year 2015&2017 and largest MSP in Israel.