Face recognition for Android using R300 model

Requirements

Runs on Android 8.0 (API level 26) or newer

Installation

Add the following dependency in your build.gradle.kts:

implementation("com.appliedred:face-recognition-r300-cloud:1.0.0")

Usage

The library requires an API key. You can obtain the API key by contacting Applied Recognition. The project’s test target includes a rate-limited API key to use for testing. You’re welcome to use the API key in your test/demo projects but it’s not suitable for production use.

The library adopts the FaceRecognition interface from the Ver-ID common types library, making it compatible with the Ver-ID SDK.

Examples

Create a face template from a face detected in an image.

import android.content.Context
import android.graphics.Bitmap
import com.appliedrec.facerecognition.r300.core.FaceTemplateVersionR300
import com.appliedrec.verid3.common.FaceTemplate
import com.appliedrec.verid3.common.Image
import com.appliedrec.verid3.common.serialization.fromBitmap
import com.appliedrec.verid3.common.use
import com.appliedrec.verid3.facedetection.retinaface.FaceDetectionRetinaFace

/**
 * Detect a face and extract its face template. Face templates are used to compare faces.
 */
suspend fun extractFaceTemplateFromFaceInImage(
    context: Context,
    bitmap: Bitmap
): FaceTemplate<FaceTemplateVersionR300, FloatArray> {
    return FaceDetectionRetinaFace.create(context).use { faceDetection ->
        val image = Image.fromBitmap(bitmap)
        faceDetection.detectFacesInImage(image, 1).firstOrNull()?.let { face ->
            FaceRecognitionR300("<your API key>", "<server URL>").use { faceRecognition ->
                faceRecognition.createFaceRecognitionTemplates(listOf(face), image)
                    .firstOrNull() ?: throw Exception("Failed to extract face template")
            }
        } ?: throw Exception("Failed to detect a face in image")
    }
}

Compare face templates.

import android.content.Context
import com.appliedrec.facerecognition.r300.core.FaceTemplateVersionR300
import com.appliedrec.verid3.common.FaceTemplate
import com.appliedrec.verid3.common.use

/**
 * Compare 2 face templates against each other and return a score ranging from 0 to 1.
 * 0 means the face templates represent 2 different people, 1 means the highest confidence
 * that the face templates represent the same person.
 */
suspend fun compareFaceTemplates(
    context: Context,
    template1: FaceTemplate<FaceTemplateVersionR300, FloatArray>,
    template2: FaceTemplate<FaceTemplateVersionR300, FloatArray>
): Float {
    return FaceRecognitionR300("<your API key>", "<server URL>").use { faceRecognition ->
        faceRecognition.compareFaceRecognitionTemplates(
            listOf(template1),
            template2
        ).firstOrNull() ?: throw Exception("Face template comparison failed")
    }
}