FaceSwap iOS

Face swap for iOS.

Background

In 2018 I decided to publish code for face swapping in Android here. I also made a face swap app for iOS.

Previous implementation depended on dlib, but it was a bit tricky to compile dlib for iOS. When Apple announced that they will support facial landmarks detection in iOS dlib was ditched. Facial landmarks are coordinates detected in a face observation. See image below for example of landmarks used. Note that all facial landmarks have not been used. Facial landmarks for eyes, mouth and nose were discarded since these will not be used by the swap algorithm. Furthermore, the landmarks coordinates for the eyebrows were adjusted slightly to include a bit of the forehead. Without adjusting those coordinates, the face swap might result in two kinds of eyebrows overlapping each other.

Dependencies

The app depends on OpenCV. It should be straightforward to add the OpenCV library to Xcode. The code that depend on OpenCV is Objective-C++ (.mm). There is a bridging-header provided not to expose C++ code to Swift.

Due to OpenCV's size, the framework was not pushed to this repository. Download it from here, in this project OpenCV-3.4.9 is used. Unzip opencv2.framework.zip, drag it to Xcode and select "Copy items if needed". The app should work with previous releases, which may reduce the app size.

To make Xcode understand that the framework exists, see the images below:

App interface

Really simple. A segment button to switch between the input photos. Tap on the screen to select photo from the camera roll. Then press the arrow to swap faces. When face swap is finished, tap the action button to share the result.

Interface

Input example

Thank you Leonardo and Brad for lending your faces.

Result

Share result

Donate

Paypal