How does DustNet Work

DustNet is a convolutional neural network (CNN), frequently referred to as an Artificial Inteligence.
We trained DustNet on millions of pieces of dust until it could solidly recognize and tag dust on its own.
When we send a newly scanned image to dustnet it sends back a map of presicely where each piece of dust is on the image. See the second example below to see the map.

Why do you even need DustNet? My slides are clean.

There's no such thing as clean film. Dust is everywhere and it loves to stick to film.
Even after applying compressed air to film dust still sticks around. Dust has been the bane of photographers since the inception of photography. Tiny pieces of dust look enormous when enlarged.

Software tools to clean up dust have never been very good. The strategy they use is to soften the entire image to blend out the dust, removing detail in the process.

DustNet takes a more precise approach. It finds the actual dust and then fixes the area under the dust only.

Look at the examples below. (Refresh the page if they don't appear) Move the slider back and forth to see the difference between the image without DustNet employed and then again with it employed. The difference is remarkable.

With and without digmypics DustNet

The image below lets you compare the dusty image to the map of the dust returned by DustNet. We then use that map to fix the image under the dust using a process called inpainting.