For example, you have a photo of two people in the foreground and a library in the background. Someone uses software or an app to remove one (or even both) of them from the photo.
What I don’t understand is how they fill in the part of the background that was originally hidden from view by the person standing in front of it. Can someone please explain this to me?
In: 7
I’ve never used AI software or anything like that, but when I do a photo edit like that it’s really just imagination, building up the scene in the context of the surroundings and trusting the viewer to not pick up on the fact that that slice of the photo doesn’t actually exist.
The original photo will have contextual clues (sky, trees, architecture, etc) but the real “trick” is the photo manipulating AI has seen many MILLIONS of photos and learned “rules” about what kind of things belong where in an image.
It can’t make something unexpected (if your friend was standing in front of a sign it can’t know what was on the sign) but trees, buildings, continuing an object that’s in other parts of the image, easy peezy
Short answer is AI. Long answer is nobody really knows. The neural network algorithm is generated by a trial and error process, and the inner workings aren’t fully understood.
Its kinda like how animal breeders were able to get desireable results before modern science.
Photoshop and a general understanding of art. If you wanna go back even further; double exposures and hand painting the negatives.
Other people in the comments are mentioning machine learning. I’m sure thats possible, but I always do it manually. Either I have a second picture with similar background elements and I layer them or I grab objects from the scene. Copy, paste, rotate, etc until it looks believable
Like others have said, it’s using AI to fill the hole. You can’t guarantee that you’ll get exactly what’s behind the object (obviously), but ideally it’s a very convincing creation based on very informed guesses based on how the AI was trained
Btw, the formal term for this is image inpainting, if you want a more in-depth info. You mentioned in another reply that you were curious who invented it. There’s a (technical) paper or two talking about the history of (digital) inpainting.
Remember when we where little and our teachers gave us coloring sheets. One of them was a grid and on one half you had say an apple on a table. The other half a blank sheet and where told to complete the picture.
From seeing the apple on a table. You assume the other half is a apple on a table fill it in.
Using photoshop or free gimp is roughly the same but with more technical knoweldge. Like blending blurry spots.
I’m no pro but I spent a solid years of my youth editing player avatars for friends using gimp and gimp tutorials. Which included cutting out charcters or background to add in different effect to image.
People who are pros tho take it to another level of detail through there knoweldge of what there doing.
Another interesting fact:
You have a spot in your eyes, where you are physically unable to see. There are tests online, where you only look at an image with one eye, and after staring long enough, a point disappears.
But your brain still fills in that information with the surrounding information. It can’t see the point anymore, and a bit around it. But it just generates a believable background.
Computers/AI/Photoshop can do the same thing as your brain.
Latest Answers