How i Used Neural Networks Data To Find The Original Colors Of A 120 Years Old Photography

Illustration : Mathieu Stern

I made a trip back through the generations to find a photograph of my great, great, great uncle, Mr Albert.
Around 1870 in France Mr Albert was an eccentric, he lived alone in a huge House away from any the nearest big city, he was e Meteorologist, a scientist and a Curiosity Collector.

Illustration : Mathieu Stern

Like many curious people in the 1870’s, Mr Albert had a keen interest in photography. My family, though, had never seen a photograph of Mr Albert himself. With his huge house now in the possession of my mother’s family, i went in search of the elusive photograph.

Amongst a box containing some broken photography equipment, i discovered three glass plates.
Two of the plates appear to be quite degraded, but one displayed a lot of promise. A hundred year old portrait of Mr Albert and his beloved dog.

To record the plate, I photographed it held up against the window and used Photoshop to get a positive Image, i then had a High definition black and white photo of my ancestor, but i didn’t want to stop there and i just heard about this new A.I. that can find original colors of a B&W photo by comparing it to billions of other photos online and calculating the aproximative variation of tones in the image, so i entered my photo in the New Neural networks algorithm and the magic happend ,i was able to find the original colors of the photo, using those datas i was able to paint them back in my photo to get a clean result.

I made a video about this experiment and i hope you will understand everything better by watching it.