Stories
Slash Boxes
Comments

SoylentNews is people

posted by Fnord666 on Thursday July 02 2020, @03:24AM   Printer-friendly
from the Sal-Maroni dept.

From TechCrunch

A new paper published by Disney Research in partnership with ETH Zurich describes a fully automated, neural network-based method for swapping faces in photos and videos — the first such method that results in high-resolution, megapixel resolution final results according, to the researchers. That could make it suited for use in film and TV, where high-resolution results are key to ensuring that the final product is good enough to reliably convince viewers as to their reality.

The researchers specifically intend this tech for use in replacing an existing actor's performance with a substitute actor's face, for instance when de-aging or increasing the age of someone, or potentially when portraying an actor who has passed away. They also suggest it could be used for replacing the faces of stunt doubles in cases where the conditions of a scene call for them to be used.

The mouse has the paper


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Thursday July 02 2020, @02:04PM (3 children)

    by Anonymous Coward on Thursday July 02 2020, @02:04PM (#1015391)

    They're not the same technique.

    Because of the GPU memory requirements and software limitations of the DeepFaceLab implementation, the highest possible resolution we were able to achieve on an 11GB GPU was 256×256 pixels. For DeepFakes we were able to produce 128×128 images. For the morphable models approach, we achieved a resolution of 500×500 pixels.

    In this work, we presented a novel approach for the unsupervised learning of multi-subject face swapping. Our method is, to our knowledge, the first to achieve convincing face-swapping results on high-resolution video in the megapixel-and-beyond domain.

  • (Score: 2) by c0lo on Thursday July 02 2020, @02:38PM (1 child)

    by c0lo (156) Subscriber Badge on Thursday July 02 2020, @02:38PM (#1015406) Journal

    The technique is the same, tho'. Note the "and software limitations of the DeepFaceLab implementation".

    --
    https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
    • (Score: 0) by Anonymous Coward on Thursday July 02 2020, @06:27PM

      by Anonymous Coward on Thursday July 02 2020, @06:27PM (#1015476)

      DeepFakes and DeepFaceLab use Y-shaped autoencoder architectures with Poisson blending, the technique in this paper uses trained Morphable Models with compositing derived from course Laplacian pyramid values in conjunction with contrast correction.

  • (Score: 0) by Anonymous Coward on Thursday July 02 2020, @03:40PM

    by Anonymous Coward on Thursday July 02 2020, @03:40PM (#1015425)

    If you train an AI to match the results of another AI... is it still the same AI?