Introduction

This face detection and alignment library provides Android software developers an easy to use method of detecting a face in a Bitmap image and obtaining the aligned face thumbnail, where eyes, nose and mouth have been standardized to fixed coordinates.

Installation

Just add the dependency to your module's gradle file:

implementation 'ai.seventhsense:facedetection:1.0.8'

Usage

First obtain an instance of a Face Detector (right now the only one we include is based on the open-source libfacedetection library).

import ai.seventhsense.facedetection.Detection;
import ai.seventhsense.facedetection.LFDDetector;
LFDDetector detector = LFDDetector.getInstance(getApplicationContext());
// You can also adjust the accuracy by specifying a higher value for maxSide
// A higher value will improve accuracy and detect smaller faces, but will 
// increase execution time. The default value for maxSide is 160, it is sufficient
// for a frontal face capture using a mobile phone front camera.
// Below we double this value:
// LFDDetector detector = LFDDetector.getInstance(getApplicationContext(), 320);

Once you have the detector instance, you can use it to detect the largest face in a Bitmap:

Bitmap bm = BitmapFactory.decodeResource(getResources(), R.drawable.tom_cruise);
Detection detection = detector.detect(bm);

The Detection object contains the co-ordinates of the bounding box of the largest face and also an aligned thumbnail which can be used with Seventh Sense' Facial Recognition algorithms.

Compiling from source

Compiling the project completely from source code is a two step process.

The first step of compiling libfacedetection can be skipped as we have already bundled the compiled binaries in this repository.

Compiling libfacedetection

Android binaries for libfacedetection are already bundled with this project and located at app/src/main/cpp/libs. You typically do not need to build them again. If you would like to proceed with the bundled binaries, please skip this section.

If you would like to build them yourself, first git clone the libfacedetection project.

Make sure you have Android Studio and NDK installed. Then, you can compile the library using cmake:

git clone https://github.com/ShiqiYu/libfacedetection
cd libfacedetection
mkdir build
cd build
cmake                                                           \
    -DCMAKE_TOOLCHAIN_FILE=/path/to/your/Android/Sdk/ndk-bundle/build/cmake/android.toolchain.cmake \
    -DANDROID_NDK=/path/to/your/Android/Sdk/ndk-bundle                               \
    -DANDROID_ABI=arm64-v8a                               \
    -DANDROID_PLATFORM=android-24                           \
    -DANDROID_STL=c++_shared                                \
    -DENABLE_NEON=ON \
    -DENABLE_AVX512=OFF \
    -DENABLE_AVX2=OFF \
    -DDEMO=OFF \
    -DUSE_OPENMP=ON ..
make

Repeat the process above with ABIs of arm64-v8a and armeabi-v7a and copy the generated libfacedetection.a to app/src/main/cpp/libs/arm64-v8a and app/src/main/cpp/libs/armeabi-v7a respectively.

Compiling this project

This project also statically links to OpenCV. Download OpenCV-android-sdk.

Unzip the OpenCV SDK and then specify its path in settings.gradle

Sync your gradle and you should be able to build the project.

License

Please see LICENSE file