The new Qualcomm Snapdragon 8 Gen 2 have a feature that will revolutionize mobile photography: semantic segmentation in real time.

Forget once about megapixels. At Andro4all we already established on one occasion that more MP does not equal better photosbasically because these only determine how big the image will be. At no time do megapixels have the responsibility of giving your photos more sharpness; only that when zooming or cropping the image it does not lose quality.

Why do we say this? Because a few months ago we already said that the new processor for the high-end, the Qualcomm Snapdragon 8 Gen 2he was going to bring with him very important changes to the field of photography. so much that would revolutionize many concepts which until now already seemed like a tremendous advance.

One of those revolutions plans unite hardware and artificial intelligence under the same umbrella, in a characteristic that from the factory in Palo Alto has come to be called semantic segmentation in images.

What is semantic segmentation in images?

What this seemingly obscure term means to the software’s ability to identify specific objects and people within a given framework. It is a central technology in many camera modes, since with it the software can identify scenes and people and then apply a certain image processing.

Many brands of smartphones already use semantic segmentation for portrait mode photos, while other brands use it to recognize scenes (as is the case with Samsung and its infamous automatic scene detection, one of the features that less favors its automatic mode, but that’s another topic). There are brands that even consider using this feature to replace elementssuch as changing a cloudy and gray sky for a blue and sunny one.

This may be another step in the right direction so that at last Mobile cameras surpass SLRssomething that Sony predicted it would happen in 2024. And the truth is that, seeing what terminals like the Google Pixel 7 Pro or the Samsung Galaxy S22 UltraTo the eye it is difficult to appreciate the differences in quality between a DSLR camera and the photos taken with these terminals.

However, Qualcomm is trying to go further and offer semantic segmentation in real time for photos and videos. The maker says its solution can identify specific things like teeth, hair, facial hair, a lip color, fabrics, and much more (you can see it in the video above), opening up an exciting world of possibilities.

What can be achieved with semantic segmentation?

Using semantic segmentation in portrait mode

The first and most obvious of the new possibilities of semantic segmentation may be a much more accurate portrait modecapable of blurring even difficult backgrounds. In the video itself, from which we have taken this capture, you can see how this technology is capable of better blurring the background leaving no focused spacessomething that still happens in most complicated situations.

From Qualcomm they have commented to different foreign media that semantic segmentation It is designed, for now, for front cameras. With semantic segmentation, they want to offer an even more precise and advanced beauty mode and, since this technology is capable of recognizing teeth, why not apply a filter that whitens them, for example? I would be just one of the many crazy things that could be achieved over time.

If we add to this that semantic segmentation can recognize hair, make it more detailed and even open the door to eliminating gray hair, manufacturers they would find themselves in a complicated moral dilemma: they would walk the fine line of embellishing photos and promoting an unrealistic standard of beauty.