This software generates a single photo that represents thousands - or more

Since photography was first invented, an estimated 2.5 trillion photos have been taken – 10% of those in the last year.

Fri, 15 Aug, 2014 - 01:00

Facebook reports 6 billion photo uploads per month and YouTube gets 72 hours of video uploaded every minute – so it wouldn’t be an exaggeration to say that we are bombarded by images on the internet.

New software developed by UC Berkeley computer scientists seeks to help us with all that, and give us one image by averaging the key features of all the other images.

How does it work?

This is an example of how the software works when typing “kids with santa” into google. You initially get something that looks like this.

((Jun-Yan Zhu, Yong Jae Lee and Alexei Efros/UC Berkeley)

That would average out into this.

The researchers weren’t happy with those results. So they proposed an interactive framework for discovering visually informative modes in the data and providing visual correspondences within each mode. If you’re still wondering, what? The below image should help clear that up.

They allowed like images to group, and by giving extra weight to specific features, the images that came out were clearer and more specific.

But what’s the point?

That’s a very good question. Some examples given by researchers include, just as an example, seeing Stephen Colbert’s typical body posture when the face of Barack Obama appears in the graphic over his shoulder. If you’re not interested in that you’re probably not a media analyst.

It could also be used to help computers. Computers need help distinguishing features in a picture because, obviously, they don’t have eyes.

When users of AverageExplorer – no that’s not internet explorer’s new name – mark eyes or a nose on an average image, the computer will automatically annotate those images on every image used to create the average image. It saves it, and subsequently you, some time.

The researchers were also inspired by James Salavon, who creates average images from hundreds of photos to illustrate a concept. For instance, how in Western cultures the bride wears white and stands to the right of the groom in photos.