Face Recognition for Ver-ID SDK using ArcFace model

Installation

1. Add the following to your project's settings.gradle.kts:

   dependencyResolutionManagement {
       repositoriesMode.set(RepositoriesMode.FAIL_ON_PROJECT_REPOS)
       repositories {
           google()
           mavenCentral()
           maven {
               url = uri("https://maven.pkg.github.com/AppliedRecognition/Ver-ID-Releases-Android")
           }
       }
   }

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

   implementation("com.appliedrec.verid3:face-recognition-arcface:1.0.0")

Usage

The library implements the FaceRecognition interface from the Ver-ID-Common-Types package. This makes it compatible with the Ver-ID SDK.

Example: Create a face template from an image and face

Use a class that implements the FaceDetection interface from the Ver-ID-Common-Types package. For example:

implementation("com.appliedrec:verid3-face-detection-retinaface:1.0.0")

To create an instance of Image import the Ver-ID serialization library by adding the following dependency:

implementation("com.appliedrec:verid3-serialization:1.0.1")

suspend fun detectFacesForRecognition(
    context: Context, 
    uri: Uri, 
    faceDetection: FaceDetection
): Array<FaceTemplateArcFace> = coroutineScope {
    // 1. Read image from URL
    val bitmap = context.contentResolver.openInputStream(uri)
        .use(BitmapFactory::decodeStream)
    // 2. Convert bitmap to Ver-ID image
    val image = Image.fromBitmap(bitmap)
    // 3. Detect up to 5 faces
    val faces = faceDetection.detectFacesInImage(image, 5)
    // 4. Create face recognition instance
    val templates = FaceRecognitionArcFace(context).use {
        // 5. Extract face templates
        it.createFaceRecognitionTemplates(faces, image) 
    }
    // 6. Return face templates
    templates
}

Example: Identify user in a face

In this example, we have a population with face templates registered for different users. We want to find the user that best matches a challenge face.

suspend fun identifyUserInFace(
   context: Context,
   challengeFace: FaceTemplateArcFace, 
   users: Array<Pair<String,FaceTemplateArcFace>>, 
   threshold: Float = 0.5f
): LinkedHashMap<String,Float> = coroutineScope {
   // 1. Create FaceRecognitionArcFace instance
   val scores = FaceRecognitionArcFace(context).use { faceRecognition ->
       // 2. Compare registered user faces to the challenge face
       faceRecognition.compareFaceRecognitionTemplates(
          users.map { it.second }.toTypedArray(), challengeFace
       )
   }
   // 3. Return users with scores matching or exceeding the threshold
   scores
      .asList() // Convert scores array to list
      .mapIndexedNotNull { index, score ->
         if (score < threshold) {
            null // Ignore element if score doesn't match the threshold
         } else {
            users[index].first to score // Return user/score pair
         }
      }
      .groupBy { it.first } // Group by user
      .map { it -> it.value.maxBy { it.second } } // Map pairs to max score for user
      .sortedByDescending { it.second } // Sort by highest score
      .toMap(LinkedHashMap()) // Convert to LinkedHashMap
}