swift – iOS Core Filter CIPersonSegmentation creates a masks whose place is completely different from unique. How one can appropriately place it?

I’ve the next picture:

I’m utilizing the beneath code to create a crimson masks from CIPersonSegmentation after which change the background coloration to cyan:

WARNING: CIPersonSegmentation requires bodily machine. It makes use of Imaginative and prescient API which I imagine shouldn’t be supported on simulator.

if let picture = UIImage(named: "demo9"), let editted = applyPersonFilter(to: picture) {
    imageView.picture = editted
}

func applyPersonFilter(to picture: UIImage) -> UIImage? {
    guard let ciImage = CIImage(picture: picture) else { return nil }
    
    print("ciImage: (ciImage.extent)") // prints (0.0, 0.0, 2882.0, 1300.0)
    
    let context = CIContext(choices: nil)
    
    let personMaskFilter = CIFilter(title: "CIPersonSegmentation", parameters: [
        kCIInputImageKey: ciImage,
        "inputQualityLevel": 0
    ])
    
    let maskImage = personMaskFilter?.outputImage?.cropped(to: ciImage.extent) as? CIImage
    
    print("maskImage: (maskImage?.extent)") // prints (0.0, 0.0, 2016.0, 1300.0)
    
    let solidColorFilter = CIFilter(title: "CIConstantColorGenerator")
    solidColorFilter?.setValue(CIColor.cyan, forKey: kCIInputColorKey)

    let blendFilter = CIFilter(title: "CIBlendWithRedMask", parameters: [kCIInputImageKey: ciImage, kCIInputBackgroundImageKey: solidColorFilter?.outputImage?.cropped(to: ciImage.extent) as Any, kCIInputMaskImageKey: maskImage as Any])
        
    guard let outputImage = blendFilter?.outputImage else {
        return nil
    }
    
    print("blendFilter: (outputImage.extent)") // prints (0.0, 0.0, 2882.0, 1300.0)
    
    guard let cgImage = context.createCGImage(outputImage, from: outputImage.extent) else { return nil }
    
    return UIImage(cgImage: cgImage)
}

The output it creates is that this:

As you may see, the masks place shouldn’t be right. It’s because the crimson masks created by CIPersonSegmentation has a distinct extent than the unique picture.

How one can appropriately apply this considering the completely different lengthen of the masks?