Adding demo for image processing

[franva]

Hi jjjkkkjjj,

Great work on bring Numpy to Swift!!!

I am learning how to do inference within CoreML using Swift.

So far I have gotten the UIImage from an image picker and I need to do preprocessing e.g. resize, transpose, normalize(mean=(0,0,0), std=(1,1,1))

And after hours and hours searching, Swift just proofed that it is not a language which is friendly for image processing. And I found your Repo here which has all the amazing feature I need.

So I think it is very helpful if you could add a demo for this.

Cheers

[jjjkkkjjj]

@franva Thank you for your feedback and sorry for late reply! I didn't suppose image processing in Matft but what Matft can connect to UIImage may be helpful. (Also complex supporting(#24)).

I found opencv's conversion codes, and referred to it. Finally, I could convert UIImage into MfArray (opposite conversion too).

The code is below;

@IBOutlet weak var originalImageView: UIImageView!
@IBOutlet weak var processedImageView: UIImageView!

func reverse(){
    let size = CFDataGetLength(self.processedImageView.image!.cgImage!.dataProvider!.data)
    let width = Int(self.processedImageView.image!.size.width)
    let height = Int(self.processedImageView.image!.size.height)

    var arr = Matft.nums(Float.zero, shape: [height, width, 4])
    var dst = Array<UInt8>(repeating: UInt8.zero, count: arr.size)

    // UIImage to MfArray
    arr.withDataUnsafeMBPtrT(datatype: Float.self){

        let srcptr = CFDataGetBytePtr(self.processedImageView.image?.cgImage?.dataProvider?.data)!

        // unit8 to float
        vDSP_vfltu8(srcptr, vDSP_Stride(1), $0.baseAddress!, vDSP_Stride(1), vDSP_Length(size))
    }

    // reverse
    arr = arr[~<<-1]
    arr = arr.conv_order(mforder: .Row)

    // MfArray to UIImage
    arr.withDataUnsafeMBPtrT(datatype: Float.self){
        srcptr in

        dst.withUnsafeMutableBufferPointer{
            // float to unit8
            vDSP_vfixu8(srcptr.baseAddress!, vDSP_Stride(1), $0.baseAddress!, vDSP_Stride(1), vDSP_Length(arr.size))

            let colorSpace = CGColorSpaceCreateDeviceRGB()
            let bitmapInfo = CGBitmapInfo(rawValue: CGImageAlphaInfo.last.rawValue | CGImageByteOrderInfo.orderDefault.rawValue)
            let provider = CGDataProvider(data: CFDataCreate(kCFAllocatorDefault, $0.baseAddress!, arr.size))
            let cgimage =  CGImage(width: arr.shape[1], height: arr.shape[0], bitsPerComponent: 8*1, bitsPerPixel: 8*arr.shape[2], bytesPerRow: arr.shape[1]*arr.shape[2], space: colorSpace, bitmapInfo: bitmapInfo, provider: provider!, decode: nil, shouldInterpolate: false, intent: CGColorRenderingIntent.defaultIntent)

            self.processedImageView.image = UIImage(cgImage: cgimage!)
        }

    }

  }

The output is like this;

Whole code is here.

Matft doesn't support full fancy indexing, so perfect image processing may not be done. but I hope this helps you.

[jjjkkkjjj]

The above code has some problems.

• Multiple conversions (UInt8 -> Float(MfArray), to Row major order, Float(MfArray) -> UInt8)
• The code is not understandable...
  • example: arr.withDataUnsafeMBPtrT(datatype: Float.self)

The second point may be solved by the next protocol version(Readme was not updated).

[jjjkkkjjj]

Hi @franva I've just implemented conversion functions (Matft.image.cgimage2mfarray and Matft.image.mfarray2cgimage) So, you can process image as mfarray!

Examples are below;

@IBOutlet weak var originalImageView: UIImageView!
@IBOutlet weak var reverseImageView: UIImageView!
@IBOutlet weak var swapImageView: UIImageView!

func reverse(){
    var image = Matft.image.cgimage2mfarray(self.reverseImageView.image!.cgImage!)

    // reverse
    image = image[Matft.reverse] // same as image[~<<-1]
    self.reverseImageView.image = UIImage(cgImage: Matft.image.mfarray2cgimage(image))
}

func swapchannel(){
    var image = Matft.image.cgimage2mfarray(self.swapImageView.image!.cgImage!)

    // swap channel
    image = image[Matft.all, Matft.all, MfArray([1,0,2,3])] // same as image[0~<, 0~<, MfArray([1,0,2,3])]
    self.swapImageView.image = UIImage(cgImage: Matft.image.mfarray2cgimage(image))
}

Please refer to the example here.