GitHub – MarekKowalski/FaceSwap: 3D face swapping implemented in Python | Photoshop Services

Share on facebook
Share on google
Share on twitter
Share on linkedin
GitHub – MarekKowalski/FaceSwap: 3D face swapping implemented in Python | Photoshop Services

FaceSwap FaceSwap is an app that I’ve initially created as an train for my college students in “Arithmetic in Multimedia” on the Warsaw College of Know-how. The app is written in Python and makes use of face alignment, Gauss Newton optimization and picture mixing to swap the face of an individual seen by the digital camera with a face of an individual in a offered picture.

One can find a brief presentation this system’s capabilities within the video under (click on to go to YouTube): click to go to YouTube

You are watching: face swap 3d

How one can use it

To start out this system you’ll have to run a Python script named photoshopservices.web (Polish for train 2). You’ll want to have Python 3 and a few further libraries put in. As soon as Python is in your machine, it is best to have the ability to mechanically set up the libraries by working pip set up -r photoshopservices.web within the repo’s root listing.

A sooner and extra secure model

A sooner and extra secure model of FaceSwap is offered on Dropbox right here. This new model is predicated on the Deep Alignment Community methodology, which is quicker than the at present used methodology if ran on a GPU and offers extra secure and extra exact facial landmarks. Please see the GitHub repository of Deep Alignment Community for setup directions.

I hope to seek out time to incorporate this sooner model within the repo code quickly.

The way it works

Do you know: best i phone photo editing app | Photoshop Services

The final define of the strategy is as follows:

First we take the enter picture (the picture of an individual we wish to see on our personal face) and discover the face area and its landmarks. As soon as we’ve got that we match the 3D mannequin to these landmarks (extra on that later) the vertices of that mannequin projected to the picture house can be our texture coordinates.

As soon as that’s completed and every part is initialized the digital camera begins capturing photos. For every captured photos the next steps are taken:

  1. The face area is detected and the facial landmarks are positioned.
  2. The 3D fashions is fitted to the positioned landmarks.
  3. The 3D fashions is rendered utilizing pygame with the feel obtained throughout initialization.
  4. The picture of the rendered mannequin is mixed with the picture obtained from the digital camera utilizing feathering (alpha mixing) and quite simple shade correction.
  5. The ultimate picture is proven to the person.

Essentially the most essential component of the complete course of is the becoming of the 3D mannequin. The mannequin itself consists of:

  • the 3D form (set of vertices) of a impartial face,
  • a lot of blendshapes that may be added to the impartial face to supply mouth opening, eyebrow elevating, and so on.,
  • a set of triplets of indices into the face form that kind the triangular mesh of the face,
  • two units of indices which set up correspondence between the landmarks discovered by the landmark localizer and the vertices of the 3D face form.

The mannequin is projected into the picture house utilizing the next equation:

Refer: 150 Best Instagram Photo Captions You Can Use – What Photography Gear

equation

the place s is the projected form, a is the scaling parameter, P are the primary two rows of a rotation matrix that rotates the 3D face form, S_0 is the impartial face form, w_1-n are the blendshape weights, S_1-n are the blendshapes, t is a 2D translation vector and n is the variety of blendshapes.

The mannequin becoming is completed by minimizing the distinction between the projected form and the localized landmarks. The minimization is completed with respect to the blendshape weights, scaling, rotation and translation, utilizing the Gauss Newton methodology.

Licensing

The code is licensed beneath the MIT license, a number of the information within the challenge is downloaded from third get together web sites:

  • brad photoshopservices.web – photoshopservices.web/wiki/Brad_Pitt#/media/File:Brad_Pitt_Fury_2014.jpg
  • photoshopservices.web – photoshopservices.web/uploads/photos/702_1433440837_albert-einstein.jpg
  • photoshopservices.web – photoshopservices.web/720×1080/a_c/Angelina-Jolie_glamour_2mar14_rex_b_720x1080.jpg
  • photoshopservices.web – photoshopservices.web/add/hands_PNG905.png
  • photoshopservices.web – photoshopservices.web/xd/521276062.jpg?v=1&c=IWSAsset&okay=2&d=62CA815BFB1CE4807BD8B4D34504661CD6D7111452E48A17257DA6DB0BD6EA6DE35742C781328F67
  • candide 3D face mannequin supply – photoshopservices.web/candide/

Contact

If need assistance otherwise you discovered the app helpful, don’t hesitate to let me know.

Marek Kowalski dichvuphotoshop.com@gmail.com, homepage: photoshopservices.web/~mkowals6/

Leave a Reply

×

Powered by WhatsApp Chat

× How can I help you?
%d bloggers like this: