Near future
Contact us
  • Home
  • Tech
  • Health
  • Environment
  • Architecture
  • energia
  • Transportation
  • Spazio
  • AI
  • concepts
  • Gadgets
  • Italy Next
  • H+
July 1 2022

Coronavirus / Russia-Ukraine

Near future

News to understand, anticipate, improve the future.

No Result
View All Result

News to understand, anticipate, improve the future.

Read in:  Chinese (Simplified)EnglishFrenchGermanItalianJapanesePortugueseRussianSpanish

Pulse, the GAN strikes again: from blurred pixels to HD photos

Pulse GAN, a neural network that with a revolutionary approach to images is able to increase the detail of blurred pixels up to 64 times as much.

Gianluca Ricciodi Gianluca Riccio
in Technology
Share39Pin11Tweet24SendShare7ShareShare5
pulse GAN
June 13 2020
⚪ Reads in 4 minutes
A A

Duke University researchers have developed an artificial intelligence tool that can turn blurry, unrecognizable images of people's faces into compelling computer-generated portraits, with finer detail than ever.

Methods prior to Pulse GAN could detail a pixelated image up to eight times its original resolution. But Duke's team devised a way to take a handful of pixels and create realistic-looking faces with up to 64 times the resolution, "imagining" features like fine lines, lashes and wrinkles that weren't in the first image. .

No images with this resolution have ever been created before

Cynthia Rudin, computer scientist, Duke University

It is not an identikit

The Pulse GAN system cannot be used to identify people, say the researchers: it will not turn a blurry and unrecognizable photo from a security camera into a crystal clear image of a real person. Rather, it is capable of generating new faces that don't exist, but seem plausibly real.

The same technique could theoretically take low-resolution photos of almost everything and create sharp and realistic images, with applications ranging from medicine and microscopy to astronomy and satellite imagery, said the co-user Sachit Menon, double specialization in mathematics and computer science.

Maybe you are also interested

Microsoft will (without haste) withdraw its technology that reads emotions

Petvation: door with facial recognition for pets

Are we ready for mind-controllable bionic body parts?

No, Google's AI is not sentient

The researchers will present their method, called PULSE GAN, from tomorrow to June 19 at the 2020 Computer Vision and Pattern Recognition (CVPR) conference.

pulse GAN, from blurred pixels to HD photos
An example of a reconstructed photo

Traditional approaches start with capturing a low-resolution image with blurry pixels and "guessing" which extra pixels are needed by trying to match them, on average, to the corresponding pixels in high-resolution images that the computer has seen before. As a result of this averaging, textured areas in hair and skin that may not line up perfectly from pixel to pixel may appear blurry and indistinct.

The Duke team has taken a different approach

Instead of capturing a low resolution image and slowly adding new details, the system looks for examples of high resolution faces generated by AI (now become very good at this), looking for the ones that look as similar as possible to the input image when reduced to the same size.

The team used a machine learning tool called GAN, or 'adversary generative network'. I have talked about it more thoroughly in this article and others on this site. GANs are neural networks trained on the same photo dataset. One network features human faces created by artificial intelligence that mimic the ones it was trained on, while the other takes this result and decides whether it's compelling enough to be mistaken for a real photo. The first network gets better and better with experience, until the second fails to distinguish. They compete with each other, in other words, and by competing they improve.

pulse GAN, from blurred pixels to HD photos

PULSE is capable of creating realistic looking images from noisy and poor quality inputs. From a single blurry image of a face it can emit an unlimited number of realistic possibilities, each of which looks subtly different.

Even given pixelated photos where the eyes and mouth are barely recognizable, “our algorithm does something about it. Something that traditional approaches cannot do ”. Word of the co-author Alex Damian, mathematician at Duke.

Pulse GAN, the "fantasy" in power

The system can convert a blurred pixel image, or a 16 × 16 pixel image to a 1024 x 1024 pixel image in seconds, adding more than a million pixels, similar to HD resolution. Details such as pores, wrinkles and strands of hair, imperceptible in low resolution photos, become sharp and clear in the computer generated versions.

The researchers asked 40 people to evaluate 1.440 images generated via PULSE GAN and five other resizing methods. Their judgment? A number on a scale of one to five. and Pulse GAN has done best of all. What's more, he scored almost equal to high-quality photos of real people.

See the results for yourself http://pulse.cs.duke.edu/.

tags: AIPhotosGANartificial intelligence
Previous post

Walkcar, the 4-wheel Ipad is on sale, and opens the era of nanomobility

Next Post

Ascento, bipedal robot but on wheels

COLLABORATE

To submit articles, disclose the results of a research or scientific discoveries write to the editorial staff

    archive

    Have a look here:

    Amphibio, wearable gills for breathing underwater
    concepts

    Amphibio, wearable gills for breathing underwater

    Amphibio is a 3D printed garment that would allow man to breathe underwater without the need for oxygen cylinders.

    Read More
    Electric vehicle emergency: the entire energy network must be designed

    Electric vehicle emergency: the entire energy network must be designed

    Inovio claims to have discovered the coronavirus vaccine

    Inovio claims to have discovered the coronavirus vaccine

    Research confirms: China knows how to successfully alter its climate

    Research confirms: China knows how to successfully alter its climate

    The robotic arm of DARPA ready in 4 years.

    The robotic arm of DARPA ready in 4 years.

    The daily tomorrow

    Futuroprossimo.it provides news on the future of technology, science and innovation: if there is something that is about to arrive, here it has already arrived. FuturoProssimo is part of the network ForwardTo, studies and skills for future scenarios.

    Subscribe to our newsletter

    Environment
    Architecture
    Artificial intelligence
    Gadgets
    concepts
    Design

    Staff
    Archives
    Advertising
    Privacy Policy

    Medicine
    Spazio
    Robotica
    Work
    Transportation
    energia

    To contact the FuturoProssimo editorial team, write to [email protected]

    Chinese Version
    Édition Française
    Deutsche Ausgabe
    Japanese version
    English Edition
    Edição Portuguesa
    Русское издание
    Spanish edition

    This work is distributed under license Creative Commons Attribution 4.0 International.
    © 2021 Futuroprossimo

    No Result
    View All Result
    • Home
    • Tech
    • Health
    • Environment
    • Architecture
    • energia
    • Transportation
    • Spazio
    • AI
    • concepts
    • Gadgets
    • Italy Next
    • H+