Face Recognition

Face recognition library for Ver-ID using ArcFace model.

Installation

Swift Package Manager

1. Open your project in Xcode.
2. Select your project in the Project Navigator.
3. Click on the Package Dependencies tab.
4. Click the + icon and enter https://github.com/AppliedRecognition/Face-Recognition-ArcFace-Apple.git in the search box labelled "Search or Enter Package URL".
5. In the Dependency Rule drop-down select Up to Next Major Version and enter 1.0.0 in the adjacent text box.
6. Press the "Add Package" button.

Usage

You will need an API key to use the face recognition in your project. The API key in tests is rate-limited and not suitable for production applications. Please contact Applied Recognition to obtain an API key.

Sample code

import UIKit
import VerIDCommonTypes
import FaceRecognitionArcFace
import FaceDetectionRetinaFaceOrt

let faceRecognition = await FaceRecognitionArcFace(apiKey: <your API key>, url: <API URL>)
let faceDetection = try await FaceDetectionRetinaFaceOrt()

func isPersonInImage(_ image1: UIImage, sameAsPersonInImage image2: UIImage) async throws -> Bool {
    // 1. Create Ver-ID images from UIImages
    guard let cgImage1 = image1.cgImage, let verIDImage1 = Image(cgImage: cgImage1) else {
        throw FaceRecognitionError.imageConversionFailed
    }
    guard let cgImage2 = image2.cgImage, let verIDImage2 = Image(cgImage: cgImage2) else {
        throw FaceRecognitionError.imageConversionFailed
    }
    
    // 2. Detect faces (assuming one face per image)
    let face1 = try await detectFaceInImage(verIDImage1)
    let face2 = try await detectFaceInImage(verIDImage2)
    
    // 3. Extract face templates from faces
    let faceTemplate1 = try await faceRecognition.createFaceRecognitionTemplates(from: [face1], in: verIDImage1).first!
    let faceTemplate2 = try await faceRecognition.createFaceRecognitionTemplates(from: [face2], in: verIDImage2).first!
    
    // 4. Compare face templates
    let score = try await faceRecognition.compareFaceRecognitionTemplates([faceTemplate1], to: faceTemplate2).first!
    
    // 5. Return whether score > threshold
    return score >= faceRecognition.defaultThreshold
}

func detectFaceInImage(_ image: Image) async throws -> Face {
    guard let face = try await faceDetection.detectFaceInImage(image, limit: 1).first else {
        throw FaceRecognitionError.noFaceDetected
    }
    return face
}

enum FaceRecognitionError: Error {
    case imageConversionFailed, noFaceDetected
}