Vision is a Framework that lets you apply high-performance image analysis and computer vision technology to images and videos, thus automatically identifying faces, detecting features and classifying scenes. It also detects barcodes, text, features and object tracking, making it an incredible tool for sorting and filtering large numbers of image files and videos.<\/p>\n
Here\u2019s what Brett Keating had to say when introducing Vision at the WWDC 2017 –<\/p>\n
\u201c<\/i><\/b><\/em><\/strong>(It is a) High-level on-device solutions to computer vision problems through one simple API<\/i><\/b><\/em><\/strong>\u201d<\/i><\/b><\/em><\/strong>.<\/i><\/b><\/em><\/strong><\/p>\n There\u2019s a whole arsenal of powerful features Vision offers. Using Core ML and deep learning, it gives you a great deal of precision and higher recall than any earlier technology. This is what enables Vision to accurately detect faces even with glasses or hats and scarves. It works exceptionally well even inside views, partially blocked faces or any other strong profiles with ease. To list out a few, here are the key capabilities you can find in Vision:<\/p>\n The simple mechanism Vision follows to obtain computer vision is:<\/p>\n Request<\/i><\/em>, Request Handler <\/i><\/em>and the Result<\/i><\/em>\u00a0of that request.<\/p>\n VNRequest<\/b><\/u><\/strong><\/p>\n This is the abstract superclass for image analysis requests. You can ask Vision to do something using this request. It has a request completion handler and an array of results.\u00a0<\/b><\/strong>i.e.<\/p>\n VNDetectFaceRectanglesRequest<\/em>\u00a0for Face detection.<\/p>\n VNDetectBarcodesRequest<\/em>\u00a0for Barcode detection.<\/p>\n VNDetectTextRectanglesRequest<\/em>\u00a0for Text region.<\/p>\n VNImageRequestHandler,<\/b><\/u><\/strong>\u00a0<\/b><\/strong>VNSequenceRequestHandler<\/b><\/u><\/strong><\/p>\n Request will execute one of the two kinds of request handlers, along with a completion block. It can execute more than one request on given image.<\/p>\n VNObservation<\/b><\/u><\/strong><\/p>\n This is the abstract superclass for image analysis results. Result array will populate with \u2018Observation\u2019.<\/p>\n Last but not the least, there is VNError<\/u><\/strong>\u00a0for error handling in Vision.<\/p>\n Best Practices<\/b><\/strong>\u00a0<\/b><\/strong><\/p>\n When introducing Vision, Apple described the following areas for Best Practices.<\/p>\n 1.\u00a0 Which image type is right for me?\u00a0 \u00a0<\/b><\/strong><\/p>\n <\/p>\n 2.\u00a0 What Am I Going to Do with the Image?<\/b><\/strong><\/p>\n <\/p>\n 3.\u00a0 What Performance Do I Need or Want?<\/b><\/strong><\/p>\n Time to get to work with Vision. Let\u2019s take a closer look at the code to see how we can detect faces from an image.<\/p>\n where handleFaceDetection<\/i><\/em>\u00a0will handle your result after performing Vision task.<\/p>\nThings you can do with Vision.<\/b><\/strong><\/h2>\n
\n
\n<\/b><\/li>\n<\/ul>\nVision Concept<\/b><\/strong><\/h2>\n
![\"Vision](\"https:\/\/www.moveoapps.com\/blog\/wp-content\/uploads\/2018\/03\/Image-1.png\")
<\/a>Before moving ahead with the code snippet, let\u2019s see the base classes and categories of\u00a0 Vision. Under the roof, there are 3 main class categories:<\/p>\n\n
\n
\n
Let<\/b><\/strong>\u2019<\/b><\/strong>s See Some Code Snippets<\/b><\/strong><\/h2>\n
\n
import\u00a0Vision<\/pre>\n
\n
var\u00a0faceDetectionRequest : VNDetectFaceRectanglesRequest\u00a0= {\r\n \r\n let\u00a0faceRequest = VNDetectFaceRectanglesRequest(completionHandler:self.handleFaceDetection)\r\n \r\n return\u00a0faceRequest\r\n \r\n}()<\/pre>\n\n
let\u00a0handler = VNImageRequestHandler(cgImage: cgImage, options: [:])\r\ndo\u00a0{\r\n try\u00a0handler.perform([self.faceDetectionRequest])\r\n} catch\u00a0{\r\n print(\"Failed to perform classification.\\n\\(error.localizedDescription)\")\r\n}\r\n<\/pre>\n<\/li>\nfunc\u00a0handleFaceDetection (request: VNRequest, error: Error?) {\r\n \r\n guard\u00a0let\u00a0observations = request.results\u00a0as? [VNFaceObservation]\r\n else\u00a0{\r\n print(\"unexpected result type from VNFaceObservation\")\r\n return\r\n }\r\n\r\n DispatchQueue.main.async\u00a0{\r\n for\u00a0face in\u00a0observations\r\n {\r\n let\u00a0view = self.createBoxView(withColor: UIColor.red)\r\n view.frame\u00a0= self.transformRect(fromRect: face.boundingBox, toViewRect: self.yourImageView)\r\n self.yourImageView.addSubview(view)\r\n }\r\n }\r\n}\r\n<\/pre>\n\n
func\u00a0transformRect(fromRect: CGRect\u00a0, toViewRect :UIView) -> CGRect\u00a0{\r\n \r\n \/\/Convert Vision Frame to UIKit Frame\r\n \r\n var\u00a0toRect = CGRect()\r\n toRect.size.width\u00a0= fromRect.size.width\u00a0* toViewRect.frame.size.width\r\n toRect.size.height\u00a0= fromRect.size.height\u00a0* toViewRect.frame.size.height\r\n toRect.origin.y\u00a0= \u00a0(toViewRect.frame.height) - (toViewRect.frame.height\u00a0* fromRect.origin.y\u00a0)\r\n toRect.origin.y\u00a0\u00a0= toRect.origin.y\u00a0- \u00a0toRect.size.height\r\n toRect.origin.x\u00a0= \u00a0fromRect.origin.x\u00a0* toViewRect.frame.size.width\r\n return\u00a0toRect\r\n \r\n}\r\n\r\nfunc\u00a0createBoxView(withColor : UIColor) -> UIView\u00a0{\r\n \r\n let\u00a0view = UIView()\r\n view.layer.borderColor\u00a0= withColor.cgColor\r\n view.layer.borderWidth\u00a0= 2\r\n view.backgroundColor\u00a0= UIColor.clear\r\n return\u00a0view\r\n}\r\n<\/pre>\n\n
![\"detecting](\"https:\/\/www.moveoapps.com\/blog\/wp-content\/uploads\/2018\/03\/Image-2.png\")
<\/a><\/p>\n![\"Facedetection](\"https:\/\/www.moveoapps.com\/blog\/wp-content\/uploads\/2018\/03\/Image-3.png\")
<\/a><\/p>\n