zjferal / android-face-landmarks

Android app that localizes facial landmarks in nearly real-time

Android App for Facial Landmark Localization

An Android Application capable of localizing a set of facial landmarks trough the frontal camera of the device with pretty-good detection performance.

Requires:

• OpenCV 4
• Dlib
• Kotlin support
• Kotlin Coroutines

Workflow

• The app utilize the frontal camera of the device to continously capture frames that are directly processed by the Face Detector built inside the Camera1 API's.
• From the detected faces only the prominent one is taken and further analized.
• The face is then converted to grayscale (OpenCV) and processed by the Dlib Face Landmark Detector.
• Finally, the localized landmarks are drawn on the UI and the entire process will repeat for the next captured frame.

Models

The models used by the Dlib landmark detector can be downloaded directly inside the app, otherwise they can be found here and here.

Usage:

1. First clone the repository.
2. Before importing the project into AndroidStudio there's a little editing to do:
   • Open the app/CMakeLists.txt
   • Then, replace with your path the variables PROJECT_PATH and OPENCV_PATH.
3. Now the project is ready to be imported into AndroidStudio.

Building Dlib from Scratch

If you want to build the latest Dlib release with custom optimization and ABIs, you can follow the instructions available here. Otherwise, you can continue with the ones that I'd already prebuilt.