Google working on AI camera tech that fixes photos before you've even taken them

Google is clearly keen on helping you get the best out of your photos and videos. Among its numerous developments in photography over the past few years, it has made an auto-white balance tool available in Google Photos, developed anti-shake fixes for videos — it has even created a great way to digitize old photos.

Now, Google is taking computational photography to the next level: it’s creating a way for you to retouch your photos, before you’ve even taken them.

Google, in cooperation with scientists at MIT, is using machine learning algorithms to improve images in your smartphone viewfinder in real-time. What’s more, these adjustments are made on on a per-case basis, rather than auto-adjustments that act the same way across every photo or scene (which many software applications already offer).

To achieve this, the team trained neural networks on 5,000 images that had been retouched by five different photographers. This gave the AI a formula from which to work from for retouching photos, resulting in more pertinent adjustments to images. Check out an example of the software in effect in the top right half of the image below and at the end of this video.

Reportedly, the software which is used to achieve this is could run in real-time on smartphones with little in the way of latency or battery life consumption — two factors which previously would have held back the implementation of such processing on phones.

This tech stands to offer a seriously worthwhile benefit to smartphone owners. Every camera has its limitations and photos can almost always be improved with a few tweaks. This software has the potential to ensure a baseline of quality is achieved, automatically, every time.

In its blog post on the subject, MIT News didn’t discuss any timeline for the potential commercialization of the software, but development seems well underway, Google is already involved, and it can run on smartphones without a significant impact on performance — I don’t see why this couldn’t appear on an Android phone in the near future.

What are your thoughts on the potential of this new tech? Let us know in the comments.