If you’ve ever scoffed at the far-fetched “Computer! Enhance!” bits in Hollywood movies and programmes such as CSI, prepare to be surprised: it’s not so far-fetched after all. Artificial intelligence is coming to your photo and video roll, and it’s going to change everything.
- 7 things we hope Google Home will be able to do in five years
- 5 great AI-powered home devices that will improve your life today
Take this photo for example. The first image is the actual photo. The final image is the result of something called Automated Texture Synthesis:
It’s not just people. Here’s a gorgeous shot of Interlaken, in Switzerland.
This one’s computer-enhanced, too. It’s a Google Street View image that’s been chucked into an experimental machine learning system. That system looks for images it likes, applies various post-processing filters and produces images almost good enough to impress professional photographers.
Take a look at these two rather good-looking people.
They don’t exist. They’re the result of an Nvidia artificial intelligence project (PDF) that uses GANs — Generative Adversarial Networks — to create portraits of entirely imaginary people. The same tech can take videos and change the weather or the season or take an image of a cat and extrapolate big cats such as lions and tigers.
This tech isn’t on your phone today, but it’s coming faster than you might expect. When artificial intelligence meets photos and videos, all kinds of interesting things happen.
Smart for art’s sake
Our smartphone cameras already do a lot of really clever stuff. They capture images before and after we press the shutter button to make sure we get the best shot. They automatically measure lighting conditions and adjust the sensor settings accordingly. They stabilise our shaky hands to keep the subject in focus.
When you start adding artificial intelligence, things get even more interesting. For example, the AI chip in the Honor View 10 smartphone — Honor calls it a NPU, which is short for Neural-Network Processing Unit — can perform facial recognition on 2,000 photos per minute, enabling you to quickly find the person you’re looking for.
That turbocharges processing that currently takes a bit of work, such as the automatic object classification both Apple and Google apply to your photo libraries — processing that means you can search for cakes as well as Claire, football as well as Fred.
When that kind of intelligence is baked into your PC or phone and tied in with other technologies such as 3D scanning, you can start doing some pretty cool things.
Think Snapchat filters that replace actual people in images and videos with photo-realistic avatars, human or otherwise, or that create realistic-looking videos of people who weren’t filmed.
Or imagine being able to say “Hey Siri, edit me a video showing my kids scoring goals at football” and get immediate results.
It’s not a huge step to that from the automated compilation videos that Apple and Google’s photo apps put to music, or that Facebook wants to show you to celebrate the fourth anniversary of being friends with Evie from accounts.
AI in everything
Over at Google, its Clips camera uses AI not just to improve photos but to decide what to shoot in the first place. For example, you might leave a Clips camera in a room with the kids while you go and do something else. When you come back, Clips will have recorded anything it thinks might be cool, because it knows who it’s looking at, where they are and what they’re up to.
So Clips might capture something like a sweet moment between the kids that you weren’t around to see, or a goal scored when you were looking the other way.
Instead of scrolling through photos or videos you’ve taken, you’ll be able to scroll through the moments Clips captured, automatically trimming, processing and sharing them with the wider world.
Adobe’s awfully excited by it all, especially when it comes to finding images. It’s using AI in what it calls Adobe Sensei, an AI system that already informs its Creative Cloud products.
AI enables Photoshop to change people’s facial expressions or the perspective of photographs without introducing distortion, it enables Premiere to create smoother edits, and stock image libraries to find photos based on ever wider criteria. And what needs a PC today will run on a phone tomorrow: remember how mad it seemed when we first saw Photoshop run on a smartphone?
AI Week is brought to you in association with Honor.
Lead image credit: Paloma Aviles, Pexels.com CC0