Package com.libremobileos.yifan.face

Class FaceFinder

java.lang.Object

com.libremobileos.yifan.face.FaceFinder

public class FaceFinder
extends Object

Combination of FaceDetector and FaceScanner for workloads where both face detection and face scanning are required. However, this class makes no assumptions about the workload and is therefore bare-bones. Because of this, usage of a task-specific class like FaceRecognizer is highly recommended, unless these do not fit your use case.

Method Summary

Modifier and Type	Method	Description
static FaceFinder	create(Context ctx, float minConfidence, int inputWidth, int inputHeight, int sensorOrientation)	Create new FaceFinder instance with sensible defaults regarding hardware acceleration (CPU, XNNPACK, 4 threads).
static FaceFinder	create(Context ctx, float minConfidence, int inputWidth, int inputHeight, int sensorOrientation, boolean hwAcceleration, boolean enhancedHwAcceleration, int numThreads)	Create new FaceFinder instance.
List<Pair<FaceDetector.Face,FaceScanner.Face>>	process(Bitmap input, boolean allowPostprocessing)	Process a Bitmap using FaceDetector, scanning the resulting found faces using FaceScanner after manually cropping the image.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

create

public static FaceFinder create(Context ctx,
                                float minConfidence,
                                int inputWidth,
                                int inputHeight,
                                int sensorOrientation,
                                boolean hwAcceleration,
                                boolean enhancedHwAcceleration,
                                int numThreads)

Create new FaceFinder instance.

Parameters:

ctx - Android Context object, may be in background.

minConfidence - Minimum confidence to track a detection, must be higher than 0.0f and smaller than 1.0f

inputWidth - width of the Bitmaps that are going to be processed

inputHeight - height of the Bitmaps that are going to be processed

sensorOrientation - rotation if the image should be rotated, or 0.

hwAcceleration - Enable hardware acceleration (NNAPI/GPU)

enhancedHwAcceleration - if hwAcceleration is enabled, use NNAPI instead of GPU. if not, this toggles XNNPACK

numThreads - How many threads to use, if running on CPU or with XNNPACK

Returns:

FaceFinder instance

See Also:

create(Context, float, int, int, int)

create

public static FaceFinder create(Context ctx,
                                float minConfidence,
                                int inputWidth,
                                int inputHeight,
                                int sensorOrientation)

Create new FaceFinder instance with sensible defaults regarding hardware acceleration (CPU, XNNPACK, 4 threads).

Parameters:

ctx - Android Context object, may be in background.

minConfidence - Minimum confidence to track a detection, must be higher than 0.0f and smaller than 1.0f

inputWidth - width of the Bitmaps that are going to be processed

inputHeight - height of the Bitmaps that are going to be processed

sensorOrientation - rotation if the image should be rotated, or 0.

Returns:

FaceFinder instance

See Also:

create(Context, float, int, int, int, boolean, boolean, int)

process

public List<Pair<FaceDetector.Face,FaceScanner.Face>> process(Bitmap input,
                                                                    boolean allowPostprocessing)

Process a Bitmap using FaceDetector, scanning the resulting found faces using FaceScanner after manually cropping the image. Adds extra metadata (location) to FaceScanner.Face based on best effort basis.

Parameters:

input - Bitmap to process.

allowPostprocessing - Allow postprocessing to improve detection quality. Undesirable when registering faces.

Returns:

List of Pairs of detection results from FaceDetector and FaceScanner