With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin

If you like to edit images in Photoshop or don't really like it, but you have to, now you will have a great new autofill feature at hand — for cases when you need to restore a damaged photo or remove an extra object from the picture.

However, you can do this not in Photoshop, but with the help of a new tool from NVIDIA. The video card manufacturer has introduced a new feature that resembles the infamous Content-Aware Autofill in Photoshop, but it works much better.

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin
Source: PetaPixel

Unlike the same function in the well-known photo editor, NVIDIA's tool does not just copy the nearest pixels to fill an empty space with them, but uses neural network technology to understand the context of the entire image and determine which missing items need to be added.

For example, in a damaged photo with a face, the neural network algorithm reads all the other elements present and understands exactly what is missing. If the eyes and nose are not visible in the photo, but there are teeth and hair, the program will understand that this is a person's face and complement it with missing organs.

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin

Let's show by example: we take the original photo and frantically move the eraser over it until we paint over most of the face.

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin

Here's what happens when you try to restore this photo in Photoshop:

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin
Autofill of a smoker

This can hardly be called the desired result. But here's what happens if you upload this photo to a program from NVIDIA:

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin

The authors of the program "trained" the neural network using more than 55 thousand randomly generated eraser strokes, shapes of holes, stripes and various other shapes, applying them to a huge number of images.

Thus, the neural network memorized the differences between damaged and original images on a large number of examples and learned how to correctly restore the missing elements.

Here are some more examples of how the program works:

With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin
With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin
With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin
With NVIDIA's new autofill tool, you will no longer paint over your eyes with skin

Keywords: Photoshop | Tools | Artificial intelligence | Computer | Neural networks | Program | Photo | Photos | Photography | Photo editor | Photoshop | Function

     

source