1. Home >
  2. Computing

Neural Networks Could Soon Bring Smartphone Photos to the Next Level

The iPhone 15 Pro Max might be impressive already, but Glass Imaging says its photos could be even better.
By Adrianna Nine
A dull, blurry, illegible photo of art references pinned to a bulletin board, next to a clearer and more saturated shot of the same art references.
Left: iPhone 15 Pro Max 5x default rendering. Right: The same shot with GlassAI applied. Credit: Glass Imaging

Smartphone cameras are pretty advanced these days, but the software attached to them prevents users from taking advantage of their full potential. That’s according to Glass Imaging, anyway: a Silicon Valley-based startup that uses artificial intelligence to “deliver DSLR quality to smartphones.” The company used the iPhone 15 Pro Max to demonstrate how Glass AI, its flagship product, can help good mobile camera hardware take even better photos.  

In its latest blog post, Glass Imaging compares regular iPhone 15 Pro Max photos with AI-boosted equivalents. In one, the button on a piece of clothing is made clearer, and the fabric around it is transformed from a garbled sea of blue into a texture-rich swatch. In another, the corner of a coffee grinder loses its noise and text artifacts, revealing intelligible labels and a smooth yet detailed exterior. Even art references pinned to a bulletin board receive a facelift through improved saturation and renewed clarity.

But how does it work? At the company’s Los Altos headquarters, the Glass Imaging team develops a series of custom neural networks for each camera built into a particular smartphone model: one for low light, one for bright light, and one for super res, or highly zoomed-in shots. (The iPhone 15 Pro Max has a main camera, an ultra-wide shooter, and a telephoto lens, totaling nine neural networks.) The team then ports those networks to the smartphone’s chipset to optimize them for efficiency. When someone snaps a photo with the modified device, Glass AI—the umbrella term for those aforementioned neural networks—receives a stack of RAW images, corrects their optical aberrations, and combines them into one processed image. 

A blurry photo of a button on a blue item of clothing next to a clearer, crisper shot of the same button/fabric.
Left: iPhone 15 Pro Max 5x default rendering. Right: The same shot with GlassAI applied. Credit: Glass Imaging
A distorted zoomed-in photo of a coffee grinder next to a clearer, readable shot of the coffee grinder's program button.
Left: iPhone 15 Pro Max 5x default rendering. Right: The same shot with GlassAI applied. Credit: Glass Imaging

The resulting photo is sharper, more saturated, and more evenly lit than an identical image garnered using only the device’s native software, the company said. It should also be an actual representation of what the photographer saw before picking up their phone, which isn’t the case with all AI-assisted photography. Glass AI isn’t making anything up, the company said; it’s just supplementing the phone’s default software to reveal its hardware’s true capabilities. 

To be clear, Glass Imaging isn’t trashing Apple’s work. Doug MacEwen, the company’s senior software engineer, notes that the iPhone 15 Pro Max’s “remarkable” 5x zoom lens “boasts impressive technical specifications.” He said it’s simply that the hardware has been “limited” by its accompanying software.

“Apple likely incorporates neural networks into some of iPhone’s image processing components,” MacEwen writes on the Glass Imaging’s blog. “However, there remains room for improvement as, in certain scenarios, the shortcomings are evident. Some neural network designs can occasionally introduce anomalies or ‘hallucinate’ details. At Glass, we place a strong emphasis on ensuring our neural networks are trained to accurately reproduce authentic details.”

Glass AI isn’t available to the public, but the company wants to partner with OEMs to implement its networks before new device releases.

Tagged In

Photography Apple IPhone Artificial Intelligence

More from Computing

Subscribe Today to get the latest ExtremeTech news delivered right to your inbox.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of use(Opens in a new window) and Privacy Policy. You may unsubscribe from the newsletter at any time.
Thanks for Signing Up