PrescreeniOS

Example

To run the example project, clone the repo, and run pod install from the Example directory first.

How-to

Initialize Prescreen instance

Prescreen.shareInstance.initialize(apiKey: "API_KEY")

Use AVCaptureVideoDataOutputSampleBufferDelegate to pass the result to scanIDCard function

extension ViewController: AVCaptureVideoDataOutputSampleBufferDelegate {
  func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
      let result = Prescreen.shareInstance.scanIDCardSync(sampleBuffer: sampleBuffer, cameraPosition: self.currentDevice.position)
      if result.error != nil {
          print(result.error!)
      }
      else if (result.confidence >= 0.5) {
          print("Confidence: \(result.confidence)")
          print("Front side: \(result.isFrontSide)")
          print("Front side full: \(String(describing: result.isFrontCardFull))")
          if (result.texts != nil) {
              print(result.texts!)
          }
          if (result.isFrontCardFull == true) {
              // available if isFrontCardFull is true
              print(result.croppedImage)
              if (result.classificationResult?.error == nil) {
                  print(result.classificationResult?.mlConfidence)
              }
              
          }
      }
      // throttle the processing 0.4s
      do { usleep(400000) }
  }
}

The IDCardResult object will consist of the following fields
```
public struct IDCardResult {
  public var error: Error?
  public var confidence: Double
  public var isFrontSide: Bool
  public var texts: [TextResult]?
  public var fullImage: UIImage?
  public var croppedImage: UIImage?
  public var isFrontCardFull: Bool?
  public var classificationResult: IDCardClassificationResult?
}
```
- error: If the recognition is successful, the error will be null. In case of unsuccessful scan, the error.errorMessage will contain the problem of the recognition.
- isFrontSide: A boolean flag indicates whether the scan found the front side (true) or back side (false) of the card ID.
- confidence: A value between 0.0 to 1.0 (higher values mean more likely to be an ID card).
- texts: A list of OCR results. An OCR result consists of type and text.
  - type: Type of information. Right now, PreScreen support 3 types
    - ID
    - SERIAL_NUMBER
    - LASER_CODE
  - text: OCR text based on the type.
- fullImage: A bitmap image of the full frame used during scanning.
- croppedImage: A bitmap image of the card. This is available if isFrontSide is true.
- isFrontCardFull: A boolean flag indicates whether the scan found thr front side and the card is likely to be complete.
- classificationResult: A result from ML classification.
  - mlConfidence: A float from 0.0 to 1.0 indicating the probability of an ID card.
  - error: A string, usually nil if no error.

Full Example

import UIKit
import AVFoundation
import PrescreeniOS

class ViewController: UIViewController {

    var backFacingCamera: AVCaptureDevice?
    var frontFacingCamera: AVCaptureDevice?
    var currentDevice: AVCaptureDevice!

    var cameraPreviewLayer: AVCaptureVideoPreviewLayer?

    let captureSession = AVCaptureSession()

    override func viewDidLoad() {
        super.viewDidLoad()
        // Do any additional setup after loading the view, typically from a nib.
        Prescreen.shareInstance.initialize(apiKey: "API_KEY")
        configure()
    }

    override func viewWillDisappear(_ animated: Bool) {
        captureSession.stopRunning()
    }

    override func didReceiveMemoryWarning() {
        super.didReceiveMemoryWarning()
        // Dispose of any resources that can be recreated.
    }

    // MARK: - Helper methods
    private func configure() {
        captureSession.sessionPreset = AVCaptureSession.Preset.hd1280x720

        let deviceDiscoverySession = AVCaptureDevice.DiscoverySession(deviceTypes: [.builtInTelephotoCamera,.builtInWideAngleCamera], mediaType: AVMediaType.video, position: .unspecified)

        for device in deviceDiscoverySession.devices {
            if device.position == .back {
                backFacingCamera = device
            } else if device.position == .front {
                frontFacingCamera = device
            }
        }

        currentDevice = backFacingCamera

        guard let captureDeviceInput = try? AVCaptureDeviceInput(device: currentDevice) else {
            return
        }

        let output = AVCaptureVideoDataOutput()
        let queue = DispatchQueue(label: "myqueue")
        output.alwaysDiscardsLateVideoFrames = true
        output.setSampleBufferDelegate(self, queue: queue)

        captureSession.addInput(captureDeviceInput)
        captureSession.addOutput(output)

        cameraPreviewLayer = AVCaptureVideoPreviewLayer(session: captureSession)
        view.layer.addSublayer(cameraPreviewLayer!)
        cameraPreviewLayer?.videoGravity = AVLayerVideoGravity.resizeAspectFill
        cameraPreviewLayer?.frame = view.layer.frame


        captureSession.startRunning()

    }
}

extension ViewController: AVCaptureVideoDataOutputSampleBufferDelegate {
    func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
        let result = Prescreen.shareInstance.scanIDCardSync(sampleBuffer: sampleBuffer, cameraPosition: self.currentDevice.position)
        if result.error != nil {
            print(result.error!)
        }
        else if (result.confidence >= 0.5) {
            print("Confidence: \(result.confidence)")
            print("Front side: \(result.isFrontSide)")
            print("Front side full: \(String(describing: result.isFrontCardFull))")
            if (result.texts != nil) {
                print(result.texts!)
            }
            if (result.isFrontCardFull == true) {
                // available if isFrontCardFull is true
                print(result.croppedImage)
                if (result.classificationResult?.error == nil) {
                    print(result.classificationResult?.mlConfidence)
                }
                
            }
        }
        // throttle the processing 0.4s
        do { usleep(400000) }
    }
}

Requirements

Installation

PrescreeniOS is available through CocoaPods. To install it, simply add the following line to your Podfile:

pod 'PrescreeniOS'

Author

northanapon, nor.thanapon@gmail.com

License

PrescreeniOS is available under the MIT license. See the LICENSE file for more info.

mistergrizzlydev / prescreen-ios-pod-1