FIRCannotCallOnMainThread - "This method must be called on a background thread." about mlkit

[bishalg]

I was able to make the code work without crashing by removing the delegate call which was on Main Thread. This was not the desired as that was the only workaround that worked for me.

For example, I have tried following style to cal the VisionFace detector on a background thread as following -

let dict: NSDictionary = [
      "imageWidth": imageWidth,
      "imageHeight": imageHeight,
      "visionImage": visionImage
    ]
    if Thread.isMainThread {
      self.performSelector(inBackground: #selector(detectFace(info:)), with: dict)
    } else {
      self.detectFacesOnDevice(in: visionImage, width: imageWidth, height: imageHeight)
    }

And then the method -

 @objc func detectFace(info: NSDictionary) {
    if let imageWidth = info["imageWidth"] as? CGFloat,
      let imageHeight = info["imageHeight"] as? CGFloat,
      let visionImage = info["visionImage"] as? VisionImage {
      self.detectFacesOnDevice(in: visionImage, width: imageWidth, height: imageHeight)
    }
  }

Even if it works well for most cases. There are cases when it doesn't work and crash on -
detectedFaces = try faceDetector.results(in: image).

For example -

    let options = VisionFaceDetectorOptions()
    //options.landmarkMode = .all
    //options.classificationMode = .all
    //options.contourMode = .all
    options.performanceMode = .accurate

Only

Performance Mode = .accurate

seems to work, other options will again trigger a crash, as the recording, affected by other Option flags.

[bcdj]

The Face detection in Firebase is deprecated. ML Kit (including face detection API) has migrated out of Firebase and become a standalone SDK. Please check out our quickstart sample app on how to hook up the face detector to a streaming video camera callback:

https://github.com/googlesamples/mlkit/blob/master/ios/quickstarts/vision/VisionExample/CameraViewController.swift#L175

It should demonstrate how to achieve what you want. Good luck!