NVIDIA has developed an algorithm that can recover the missing pieces of the photographs, according to Engadget. It works on the basis of deep learning.
The NVIDIA tool analyzes the image and based on their own knowledge to decide what can be on the location of the missing fragments. For example, if the portrait is lost one eye, the algorithm realizes that it’s face and scratch should be the eye and not the other part of the body.
To NVIDIA, other companies attempted to create a similar tool. But their development is simply copying adjacent pixels. Algorithm NVIDIA does not depend on the initial data. The company claims that it can work with empty fragments of any size and shape.
To train the algorithm, needed to create tens of thousands of variations of the disadvantages. During testing, the algorithm just gave a blank fragments from other images, so he knew how to recover photos.
The results of the tool NVIDIA is not always perfect. For example, the facial features, which is shown in the video was clearly taken from another image. But it all looks plausible.
Follow the news in our Telegram channel (if you can bypass the lock), as well as in the app on iOS MacDigger.