How is it possible to use Android Camera APIs in non-Camera applications?

Question

I would like to use some Android 4 APIs in a non-Camera application.

The API includes some very nice Face Detection classes, including the Camera.Face class available since API 14. I would like to apply the same Face Detection classes in order to implement face detection on images saved on the device. I would prefer to use this to process pictures stored on the device itself (ex: social tagging, face image manipulation, etc.)

I require guidance on how to accomplish this re-use task.

Answer 1

If what you need is to detect faces in images stored on the device, you can definitely do this without hacking the source code of android!

There is a FaceDetector API that is available under the package android.media since API 1, which accepts Bitmap as input (formatted in 565 format) and give you the position of faces in that picture.

Here are the steps you need:

1- Load the Bitmap and convert it to 565 format (assuming you have facesPicture file under your drawable resources)

Bitmap originalBitmap = 
            BitmapFactory.decodeResource(getResources(),R.drawable.facesPicture);

Bitmap bitmap = originalBitmap .copy(Bitmap.Config.RGB_565, true);

originalBitmap .recycle(); // allow the GC to collect this object

2- Define Face array to hold the detected faces information and initialize the FaceDetector

int MAX_FACES = 20; // assuming that the image will have maximum 20 faces

FaceDetector.Face[] faces = new FaceDetector.Face[MAX_FACES];

FaceDetector faceDetector = 
             new FaceDetector(bitmap.getWidth(), bitmap.getHeight(), MAX_FACES);

3- Search for the faces and process results

int facesCount = faceDetector.findFaces(bitmap, faces);

for(int i=0; i<facesCount; i++) {
    FaceDetector.Face face = faces[i];

    float detectionConfidence = face.confidence(); // over 0.3 is OK

    PointF eyesMidPoint = new PointF();

    face.getMidPoint(eyesMidPoint);

    float eyesDistance = face.eyesDistance();

    float rotationX = face.pose(FaceDetector.Face.EULER_X);

    float rotationY = face.pose(FaceDetector.Face.EULER_Y);

    float rotationZ = face.pose(FaceDetector.Face.EULER_Z);

    // Do something with these values

}

You can download a complete project example here which is explained in this article Face Detection with Android APIs

If you want something more advanced you should consider using OpenCV