CreateML图像分类模型在iOS应用中输出错误结果求助

我使用CreateML创建了一个图像分类模型，可将图像分为猫、狗、兔子三类。在iOS应用中，点击imageView从相册选择图像，由该视觉模型处理后将结果显示在Label上。

在CreateML中测试模型时一切正常，但在应用中，每次启动后无论选择什么图像，都会依次输出错误结果：

• 77% Dog
• 83% Cat
• 71% Cat
• 56% Cat ...

我已尝试以下方法但问题未解决：

• 在不同线程执行操作
• 调整所选图像尺寸
• 重新制作模型
• 修改PHPicker配置

应用相关截图：

• 应用首次打开界面
• 选择相册图像界面
• 首次输出结果
• 第二次输出结果

相关代码 #

import UIKit
import Photos
import PhotosUI
import CoreML
import Vision

class ViewController: UIViewController, PHPickerViewControllerDelegate, UINavigationBarDelegate {
    
    @IBOutlet weak var imageView: UIImageView!
    
    @IBOutlet weak var resultLabel: UILabel!
    
    override func viewDidLoad() {
        super.viewDidLoad()
        
        imageView.isUserInteractionEnabled = true
        
        let gesture = UITapGestureRecognizer(target: self, action: #selector(selectImage))
        
        imageView.addGestureRecognizer(gesture)
    }
    
    @objc func selectImage(){
        
        imageView.image = UIImage()
        var configuration = PHPickerConfiguration(photoLibrary: .shared())
        
        configuration.selectionLimit = 1
        configuration.filter = PHPickerFilter.images
        let vc = PHPickerViewController(configuration: configuration)
        vc.delegate = self
        present(vc, animated: true)
    }
    
    func picker(_ picker: PHPickerViewController, didFinishPicking results: [PHPickerResult]) {
        picker.dismiss(animated: true)
        
        DispatchQueue.global().async {
            results[0].itemProvider.loadObject(ofClass: UIImage.self) { [weak self] (reading, error) in
                guard var imageSelected = reading as? UIImage, error == nil else {return}
                print("Selected image = \(imageSelected)")
                
                guard let ciimage = CIImage(image: imageSelected) else {fatalError("Problem while converting to CIImage")}
                
                self?.detectImage(image: ciimage)
                
                
                DispatchQueue.main.async {
                    self?.imageView.image = imageSelected
                }
            }
        }
    }
    
    func detectImage(image : CIImage){
        
        let config = MLModelConfiguration()
        
        guard let model  = try? VNCoreMLModel(for: DogCatRabbitMLTry_1.init(configuration: config).model) else {fatalError("Loading CoreML Model Failed")}
        
        let request = VNCoreMLRequest(model: model) { (request, error) in
            guard let results = request.results as? [VNClassificationObservation] else {
                    fatalError("Model failed to process image")
            }
            if let firstResult = results.first{
                print(firstResult.identifier)
                
                DispatchQueue.main.async {
                    self.resultLabel.text = "%\(Int(firstResult.confidence*100)) \(firstResult.identifier)"
                }
            }
        }
        
        let handler = VNImageRequestHandler(ciImage: image)
        
        do{
            try handler.perform([request])
        } catch{
            fatalError("ciimage was not handling")
        }
    }
    
    /*
    func resizeImage(image: UIImage, targetSize: CGSize) -> UIImage? {
        let size = image.size
        
        let widthRatio  = targetSize.width  / size.width
        let heightRatio = targetSize.height / size.height
        
        // Figure out what our orientation is, and use that to form the rectangle
        var newSize: CGSize
        if(widthRatio > heightRatio) {
            newSize = CGSize(width: size.width * heightRatio, height: size.height * heightRatio)
        } else {
            newSize = CGSize(width: size.width * widthRatio, height: size.height * widthRatio)
        }
        
        // This is the rect that we've calculated out and this is what is actually used below
        let rect = CGRect(origin: .zero, size: newSize)
        
        // Actually do the resizing to the rect using the ImageContext stuff
        UIGraphicsBeginImageContextWithOptions(newSize, false, 1.0)
        image.draw(in: rect)
        let newImage = UIGraphicsGetImageFromCurrentImageContext()
        UIGraphicsEndImageContext()
        
        return newImage
    }
     */
}

内容的提问来源于stack exchange，提问作者Muammer Akca