//
//  ViewController.swift
//  ML
//
//  Created by MoneyLee on 2020/8/21.
//  Copyright © 2020 MoneyLee. All rights reserved.
//

import UIKit
//机器学习必须引用的
import CoreML
//视界 图像类的工具
import Vision

class ViewController: UIViewController {

    @IBOutlet weak var showImgV: UIImageView!
    
    override func viewDidLoad() {
        super.viewDidLoad()
        // Do any additional setup after loading the view.
    }

    @IBAction func camreClick(_ sender: UIBarButtonItem) {
        let picker = UIImagePickerController()
        picker.delegate = self
        picker.sourceType = .camera
        present(picker, animated: true, completion: nil)
    }
}

extension ViewController: UIImagePickerControllerDelegate, UINavigationControllerDelegate {
    
    func imagePickerController(_ picker: UIImagePickerController, didFinishPickingMediaWithInfo info: [UIImagePickerController.InfoKey : Any]) {
        
        if let img = info[.originalImage] as? UIImage {
            showImgV.image = img
            navigationItem.title = "识别中..."
            //转化为CIImage
            guard let ciImg = CIImage(image: img) else {
                fatalError("不能把图像转化为CIImage")
            }
            //加载训练好的模型
            guard let model = try? VNCoreMLModel(for: MobileNetV2FP16().model) else {
                fatalError("加载ML Model失败")
            }
            
            //创建识别请求
            let request = VNCoreMLRequest(model: model) { (req, error) in
                //请求的回调 res 识别的结果
                guard let res = req.results else {
//                    fatalError("图像识别失败")
                    self.navigationItem.title = "图片识别失败"
                    return
                }
                
                let classifcations = res as! [VNClassificationObservation]
                if classifcations.isEmpty {
                    print("没有识别成功")
                    self.navigationItem.title = "没有识别成功"
                } else {
                    //取识别概率最大的结果
                    print(classifcations.first!.identifier)
                    self.navigationItem.title = classifcations.first!.identifier
                }
                
            }
            //苹果的model图像是正方型, 我们将获取的图像转换为正方型 有助于识别, 默认是原图
            request.imageCropAndScaleOption = .centerCrop
            
            //执行识别请求
            do {
                try VNImageRequestHandler(ciImage: ciImg).perform([request])
            } catch {
                print("执行图像识别请求失败, 原因是: \(error.localizedDescription)")
            }
        }
        picker.dismiss(animated: true, completion: nil)   
    }
}

coremltools.converters.caffe.convert(model, image_input_names=[], class_labels=None, predicted_feature_name=None, model_precision='float32')

Python 3.7.9 (v3.7.9:13c94747c7, Aug 15 2020, 01:31:08) 
[Clang 6.0 (clang-600.0.57)] on darwin
>>> import coremltools
>>> 
>>> coreml_model = coremltools.converters.caffe.convert(('oxford102.caffemodel','deploy.prototxt'), image_input_names='data', class_labels='class_labels.txt')

================= Starting Conversion from Caffe to CoreML ======================
Layer 0: Type: 'Input', Name: 'input'. Output(s): 'data'.
Ignoring batch size and retaining only the trailing 3 dimensions for conversion. 
Layer 1: Type: 'Convolution', Name: 'conv1'. Input(s): 'data'. Output(s): 'conv1'.
Layer 2: Type: 'ReLU', Name: 'relu1'. Input(s): 'conv1'. Output(s): 'conv1'.
Layer 3: Type: 'Pooling', Name: 'pool1'. Input(s): 'conv1'. Output(s): 'pool1'.
Layer 4: Type: 'LRN', Name: 'norm1'. Input(s): 'pool1'. Output(s): 'norm1'.
Layer 5: Type: 'Convolution', Name: 'conv2'. Input(s): 'norm1'. Output(s): 'conv2'.
Layer 6: Type: 'ReLU', Name: 'relu2'. Input(s): 'conv2'. Output(s): 'conv2'.

================= Summary of the conversion: ===================================
Detected input(s) and shape(s) (ignoring batch size):
'data' : 3, 227, 227

Network Input name(s): 'data'.
Network Output name(s): 'prob'.

>>>

>>> coreml_model.save('oxford102.mlmodel')

链接:https://pan.baidu.com/s/1_WXy4jiTOkusapYv37S3rw  密码:8bvu

链接:https://pan.baidu.com/s/1aUDT5HbXgjEOmAuOluY9DA  密码:rvae

import Cocoa
import CreateML

//总数据
let data = try MLDataTable(contentsOf: URL(fileURLWithPath: "/Users/moneylee/Documents/Temp/spam.csv"))

//把总数据分为训练数据和测试数据
let (trainingData,testingData) = data.randomSplit(by: 0.8, seed: 5)

//开始训练--并得出情感识别器（sentimentClassifier）-也就是训练出来的mlmodel
let sentimentClassifier = try MLTextClassifier(trainingData: trainingData, textColumn: "text", labelColumn: "label")


//设置mlmodel元数据
let metadata = MLModelMetadata(author: "Lebus", shortDescription: "A model trained to classify spam SMS", version: "1.0")
//保持mlmodel
try sentimentClassifier.write(to: URL(fileURLWithPath: "/Users/moneylee/Documents/Temp/spam.mlmodel"), metadata: metadata)

//识别是否是垃圾短信
try sentimentClassifier.prediction(from: "hi,how are you?")
try sentimentClassifier.prediction(from: "Had your mobile 11 months or more? U R entitled to Update to the latest colour mobiles with camera for Free! Call The Mobile Update Co FREE on 08002986030")

链接:https://pan.baidu.com/s/1axNkcjwNE4bkpFPFJXYhLQ  密码:ubti