Google And MIT To Use AI To Retouch Photos Before They’re Captured

When improving the camera’s hardware performance won’t make much of a difference to the consumer (a thing we’ve already started to notice), phone manufacturers will have to find other ways to tempt users. Like Google is doing right now with the help of MIT researchers. By training neural networks, the tech behemoth believes it can enable the phone’s camera to retouch pictures before they’re captured #machinemagic

MIT engineers believe they can teach a smartphone’s camera to edit the “ideal” picture in a given setting, just like a professional photographer learned how to enhance a fashion shot or wildlife pic. The key ingredient is machine learning. 

They have already trained neural networks with a set of 5,000 images. Each image was given to five separate photographers, invited by the companies to retouch them as they feel like. After the first editing round, the algorithms broke down the process and learned what changes need to be made to deliver a quality photo. Before you leave this article unimpressed, let us tell you this – MIT and Google are pretty confident that their AI tech is capable of applying the improvements in real time, without any lag, before capture. 

Imagine getting your phone to reproduce a photographer’s unique style in camera, before pressing the shutter button, without needing him to actually be there. That’s pretty awesome, right? Given Google’s new and successful smartphone endeavor, we could actually see the tech in action, on one of their Pixel devices.

When can we expect such a feature? Well, given the fact that the tech would have to run continuously, from the moment you open the camera, the phone would need unprecedented power. That’s not impossible, though, just pretty difficult to achieve without dedicated hardware. We’re talking about an imaging engine with shorter execution times and the lowest power consumption possible like FotoNation‘s IPU. As they say, “it is not enough to have a high quality software implementation if after 10 minutes of taking photos users can’t answer a call because their phone is burning up.”