Why isn't my face detection code using CIDetector working properly?

• The first face isns't detected in my app, only in Camera.app.
• For the third face — the east asian woman — Camera.app correctly draws a rectangle around her face, while my app draws a rectangle that extends far below her face.
• Obama's face isn't detected in my app, only in Camera.app.
• When the camera zooms out from Putin's face, my app draws a rectangle over the right half of his face, cutting it in half, while Camera.app draws a rectangle correctly around his face.

Why is this happening?

My code is as follows. Do you see anything wrong?

First, I create a video output as follows:

let videoOutput = AVCaptureVideoDataOutput()
videoOutput.videoSettings =
    [kCVPixelBufferPixelFormatTypeKey as AnyHashable:
    Int(kCMPixelFormat_32BGRA)]
session.addOutput(videoOutput)
videoOutput.setSampleBufferDelegate(faceDetector, queue: faceDetectionQueue)

This is the delegate:

class FaceDetector: NSObject, AVCaptureVideoDataOutputSampleBufferDelegate {
  func captureOutput(_ captureOutput: AVCaptureOutput!,
                     didOutputSampleBuffer sampleBuffer: CMSampleBuffer!,
                     from connection: AVCaptureConnection!) {
    let imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer)!
    let features = FaceDetector.ciDetector.features(
        in: CIImage(cvPixelBuffer: imageBuffer))


    let faces = features.map { $0.bounds }
    let imageSize = CVImageBufferGetDisplaySize(imageBuffer)


    let faceBounds = faces.map { (face: CIFeature) -> CGRect in
        var ciBounds = face.bounds


        ciBounds = ciBounds.applying(
            CGAffineTransform(scaleX: 1/imageSize.width, y: -1/imageSize.height))
        CGRect(x: 0, y: 0, width: 1, height: -1).verifyContains(ciBounds)


        let bounds = ciBounds.applying(CGAffineTransform(translationX: 0, y: 1.0))
        CGRect(x: 0, y: 0, width: 1, height: 1).verifyContains(bounds)
        return bounds
    }
    DispatchQueue.main.sync {
      facesUpdated(faceBounds, imageSize)
    }
  }

  private static let ciDetector = CIDetector(ofType: CIDetectorTypeFace,
      context: nil,
      options: [CIDetectorAccuracy: CIDetectorAccuracyHigh])!
}

The facesUpdated() callback is as follows:

class PreviewView: UIView {
  private var faceRects = [UIView]()


  private static func makeFaceRect() -> UIView {
    let r = UIView()
    r.layer.borderWidth = FocusRect.borderWidth
    r.layer.borderColor = FocusRect.color.cgColor
    faceRects.append(r)
    addSubview(r)
    return r
  }


  private func removeAllFaceRects() {
    for faceRect in faceRects {
      verify(faceRect.superview == self)
      faceRect.removeFromSuperview()
    }
    faceRects.removeAll()
  }


  private func facesUpdated(_ faces: [CGRect], _ imageSize: CGSize) {
    removeAllFaceRects()


    let faceFrames = faces.map { (original: CGRect) -> CGRect in
        let face = original.applying(CGAffineTransform(scaleX: bounds.width, y: bounds.height))
        verify(self.bounds.contains(face))
        return face
    }


    for faceFrame in faceFrames {
      let faceRect = PreviewView.makeFaceRect()
      faceRect.frame = faceFrame
    }
  }
}

I also tried the following, but they didn't help:

• Setting the AVCaptureVideoDataOutput's videoSettings to nil.
• Explicitly setting the CIDetector's orientation to portrait. The phone is in portrait for this test, so it shouldn't matter.
• Setting and removing CIDetectorTracking: true
• Setting and removing CIDetectorAccuracy: CIDetectorAccuracyHigh
• Trying to track only one face, by looking only at the first feature detected.
• Replacing CVImageBufferGetDisplaySize() with CVImageBufferGetEncodedSize() — they're anyway same, at 1440 x 1080.