Vision Framework

PDF for offline use
Sample Code:
Related Videos:

Let us know how you feel about this

Translation Quality


0/250

last updated: 2016-08

The Vision framework adds a number of new image processing features to iOS 11, including:

Photograph with three rectangles detected Photograph with two faces detected

Rectangle Detection and Face Detection are discussed in more detail below.

Rectangle Detection

The VisionRects sample shows how to process an image and draw the detected rectangles on it.

1. Initialize the Vision request

In ViewDidLoad, create a VNDetectRectanglesRequest that references the HandleRectangles method which will be called at the end of each request:

The MaximumObservations property should also be set, otherwise it will default to 1 and only a single result will be returned.

RectangleRequest = new VNDetectRectanglesRequest(HandleRectangles);
RectangleRequest.MaximumObservations = 10;

2. Start the Vision processing

The following code starts processing the request. In the VisionRects sample, this code runs after the user has selected an image:

// Run the rectangle detector
var handler = new VNImageRequestHandler(ciImage, uiImage.Orientation.ToCGImagePropertyOrientation(), new VNImageOptions());
DispatchQueue.DefaultGlobalQueue.DispatchAsync(()=>{
  handler.Perform(new VNRequest[] {RectangleRequest}, out NSError error);
});

This handler passes the ciImage to the Vision framework VNDetectRectanglesRequest that was created in step 1.

3. Handle the results of Vision processing

Once the rectangle detection is complete, the framework executes the HandleRectangles method, a summary of which is shown below:

private void HandleRectangles(VNRequest request, NSError error){
  var observations = request.GetResults<VNRectangleObservation>();
  // ... omitted error handling ...
  bool atLeastOneValid = false;
  foreach (var o in observations){
    if (InputImage.Extent.Contains(boundingBox)) {
      atLeastOneValid |= true;
    }
  }
  if (!atLeastOneValid) return;
  // Show the pre-processed image
  DispatchQueue.MainQueue.DispatchAsync(() =>
  {
    ClassificationLabel.Text = summary;
    ImageView.Image = OverlayRectangles(RawImage, imageSize, observations);
  });
}

4. Display the results

The OverlayRectangles method in the VisionRectangles sample has three functions:

  • Rendering the source image,
  • Drawing a rectangle to indicate where each one was detected, and
  • Adding a text label for each rectangle using CoreGraphics.

View the sample's source for the exact CoreGraphics method.

Photograph with three rectangles detected

5. Further processing

Rectangle detection is often just the first step in a chain of operations, such as with this CoreMLVision example, where the rectangles are passed to a CoreML model to parse handwritten digits.

Face Detection

The VisionFaces sample works in a similar fashion to the VisionRectangles sample, using a different Vision request class.

1. Initialize the Vision request

In ViewDidLoad, create a VNDetectFaceRectanglesRequest that references the HandleRectangles method which will be called at the end of each request.

FaceRectangleRequest = new VNDetectFaceRectanglesRequest(HandleRectangles);

2. Start the Vision processing

The following code starts processing the request. In the VisionFaces sample this code runs after the user has selected an image:

// Run the face detector
var handler = new VNImageRequestHandler(ciImage, uiImage.Orientation.ToCGImagePropertyOrientation(), new VNImageOptions());
DispatchQueue.DefaultGlobalQueue.DispatchAsync(()=>{
  handler.Perform(new VNRequest[] {FaceRectangleRequest}, out NSError error);
});

This handler passes the ciImage to the Vision framework VNDetectFaceRectanglesRequest that was created in step 1.

3. Handle the results of Vision processing

Once the face detection is complete, the handler executes the HandleRectangles method which performs error handling and displays the bounds of the detected faces, and calls the OverlayRectangles to draw bounding rectangles on the original picture:

private void HandleRectangles(VNRequest request, NSError error){
  var observations = request.GetResults<VNFaceObservation>();
  // ... omitted error handling...
  var summary = "";
  var imageSize = InputImage.Extent.Size;
  bool atLeastOneValid = false;
  Console.WriteLine("Faces:");
  summary += "Faces:";
  foreach (var o in observations) {
    // Verify detected rectangle is valid. omitted
    var boundingBox = o.BoundingBox.Scaled(imageSize);
    if (InputImage.Extent.Contains(boundingBox)) {
      atLeastOneValid |= true;
    }
  }
  // Show the pre-processed image (on UI thread)
  DispatchQueue.MainQueue.DispatchAsync(() =>
  {
    ClassificationLabel.Text = summary;
    ImageView.Image = OverlayRectangles(RawImage, imageSize, observations);
  });
}

4. Display the results

The OverlayRectangles method in the VisionFaces sample has three functions:

  • Rendering the source image,
  • Drawing a rectangle for each face detected, and
  • Adding a text label for each face using CoreGraphics.

View the sample's source for the exact CoreGraphics method.

Photograph with two faces detected

5. Further processing

The Vision framework includes additional capabilities to detect facial features, such as the eyes and mouth. Use the VNDetectFaceLandmarksRequest type, which will return VNFaceObservation results as in step 3 above, but with additional VNFaceLandmark data.

Xamarin Workbook

If it's not already installed, install the Xamarin Workbooks app first. The workbook file should download automatically, but if it doesn't, just click to start the workbook download manually.