Check out Google RAISR: Sharp images with machine learning

If you share a picture of a tree in a forest, but no one can see it, did you really share it?

Working at Google, where teams aspire to “three-comma moments” (i.e. reaching 1,000,000,000 users), it’s become overwhelmingly clear to me that all the fancy features in the world don’t mean squat if people can’t access them. And traveling in Nepal, I got a taste of just how slow & expensive connectivity can be. Anything that helps deliver content faster & more cheaply means more democratic access to ideas & inspiration.

That’s why I’ve been really excited about RAISR (“Rapid and Accurate Image Super-Resolution”). The Google Research team writes,

[It’s] a technique that incorporates machine learning in order to produce high-quality versions of low-resolution images. RAISR produces results that are comparable to or better than the currently available super-resolution methods, and does so roughly 10 to 100 times faster, allowing it to be run on a typical mobile device in real-time.


I’ve been championing this tech within the company and—because the research paper is public—encouraging friends at Facebook, Adobe, Apple, and elsewhere to check it out. Fast, affordable access is good for everyone.

It’s funny: I came here to “teach Google Photoshop” (i.e. to make computers see & create like artists), yet if I do my job right here, you’ll never spot a thing. I’ve come to prioritize access far ahead of synthesis. Funny ol’ world.

PS—Obligatory (?) Old Man Nack remark: “In my day, Genuine Fractals, blah blah…”

One thought on “Check out Google RAISR: Sharp images with machine learning

  1. Looks awesome John…. and from the above sample doesn’t look “over” sharpened which a lot of digital work does….

Leave a Reply

Your email address will not be published. Required fields are marked *