Is there a way to make MLVision text recognition faster?

Question

I am using MLVision cloud text recognition for my app. I capture/upload a photo and then I start the process. When it recognises the image and extract the text, then I separate it and append every separated block into an array.

The code below is for the whole process.

    lazy var vision = Vision.vision()
    var textRecognizer: VisionTextRecognizer!
    var test = [] as Array<String>



override func viewDidLoad() {
        super.viewDidLoad()

        let options = VisionCloudTextRecognizerOptions()
        options.languageHints = ["en","hi"]
        textRecognizer = vision.cloudTextRecognizer(options: options)
    }


//where pickedImage is the image that user captures.
let visionImage = VisionImage(image: pickedImage)
            textRecognizer.process(visionImage, completion: { (features, error) in
                guard error == nil, let features = features else {
                    self.resultView.text = "Could not recognize any text"
                    self.dismiss(animated: true, completion: nil)
                    return
                }

                for block in features.blocks {
                    for line in block.lines{

                        //for element in line.elements{
                        self.resultView.text = self.resultView.text + "\(line.text)"

                    }
                }
                self.separate()

            })

func separate(){
        let separators = CharacterSet(charactersIn: (":)(,•/·]["))
        let ofWordsArray = self.resultView.text.components(separatedBy: separators)

        for word in ofWordsArray{

            let low = word.trimmingCharacters(in: .whitespacesAndNewlines).lowercased()
            if low != ""{
            test.append(low)
            }


        }
        print(test)
    }

Everything works fine and I get the result that I want.The problem is that I think is really slow. It takes about 20sec for the entire process.Is there a way to make it faster? Thanks in advance.

Answer 1

You are using the VisionCloudTextRecognizer. Speed will depend on your connection, in my case it was only few seconds. Your other option is to use on-device text recognition or use a hybrid approach, where you first detect on-device, then correct with Cloud API later.