【iOS】调用实时摄像头图像识别

最新推荐文章于 2023-11-26 15:50:57 发布

RodrickOMG

最新推荐文章于 2023-11-26 15:50:57 发布

阅读量1.8k

点赞数

分类专栏： iOS 文章标签： iOS Core ML Models swift xcode 图像识别

本文链接：https://blog.csdn.net/daiyucheng88/article/details/99559356

版权

iOS 专栏收录该内容

5 篇文章

订阅专栏

本文介绍了一种小型深度神经网络SqueezeNet，该网络能够实现实时的图像分类，通过使用苹果公司提供的模型，结合AVFoundation框架进行摄像头画面捕获，并利用CoreML和Vision框架进行图像识别。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

SqueezeNet

A small Deep Neural Network architecture that classifies the dominant object in a camera frame or image.

模型下载：https://developer.apple.com/machine-learning/models/

override func viewDidLoad() {
        super.viewDidLoad()
        
		let captureSession = AVCaptureSession()
        captureSession.sessionPreset = .photo
		//AVCaptureSession 继承自NSObject，是AVFoundation的核心类；用于管理捕获对象AVCaptureInput的视频和音频的输入，协调捕获的输出AVCaptureOutput

        guard let captureDevice = AVCaptureDevice.default(.builtInWideAngleCamera, for: .video, position: .front) else { return }
        guard let input = try? AVCaptureDeviceInput(device: captureDevice) else { return }
        //guard let守护一定有值。如果没有值，直接返回。通常判断是否有值之后，会做具体的逻辑实现
        captureSession.addInput(input)
        //添加输入
        
        captureSession.startRunning()
        
        let previewLayer = AVCaptureVideoPreviewLayer(session: captureSession)
        view.layer.addSublayer(previewLayer)
        previewLayer.videoGravity = AVLayerVideoGravity.resizeAspectFill
        previewLayer.frame = CGRect(x: 20, y: 270, width: screenWidth - 40, height: 400)
        //设定捕捉摄像头的画面layout属性和大小
        
        let dataOutput = AVCaptureVideoDataOutput()
        dataOutput.setSampleBufferDelegate(self, queue: DispatchQueue(label: "videoQueue"))
        captureSession.addOutput(dataOutput)
        //将捕捉到的画面输出
        
        setupIdentifierConfidenceLabel()
        
        
//        let request = VNCoreMLRequest(model: <#T##VNCoreMLModel#>, completionHandler: <#T##VNRequestCompletionHandler?##VNRequestCompletionHandler?##(VNRequest, Error?) -> Void#>)
//        VNImageRequestHandler(cgImage: <#T##CGImage#>, options: [:]).perform(<#T##requests: [VNRequest]##[VNRequest]#>)
        
    }

 	private func setupIdentifierConfidenceLabel() {
        view.addSubview(identifierLabel)
        identifierLabel.bottomAnchor.constraint(equalTo: view.bottomAnchor, constant: -32).isActive = true
        identifierLabel.leftAnchor.constraint(equalTo: view.leftAnchor).isActive = true
        identifierLabel.rightAnchor.constraint(equalTo: view.rightAnchor).isActive = true
        identifierLabel.heightAnchor.constraint(equalToConstant: 50).isActive = true
    }
    
    
    func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
        //        print("Camera was able to capture a frame:", Date())
        
        guard let pixelBuffer: CVPixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else { return }
        
        // !!!Important
        // make sure to go download the models at https://developer.apple.com/machine-learning/ scroll to the bottom
        guard let model = try? VNCoreMLModel(for: SqueezeNet().model) else { return }
        let request = VNCoreMLRequest(model: model) { (finishedReq, err) in
            
            //perhaps check the err
            
            //            print(finishedReq.results)
            
            guard let results = finishedReq.results as? [VNClassificationObservation] else { return }
            
            guard let firstObservation = results.first else { return }
            
            print(firstObservation.identifier, firstObservation.confidence)
            
            DispatchQueue.main.async {
                self.identifierLabel.text = "\(firstObservation.identifier) \(firstObservation.confidence * 100)"//显示图像识别内容以及确信度
            }
        }
        
        try? VNImageRequestHandler(cvPixelBuffer: pixelBuffer, options: [:]).perform([request])
    }