Using SkiaSharp for imageing but not System.Drawing.Common for linux support

[bnuzhouwei]

System.Drawing is not suit for linux, for libgdiplus has memory leak problem in linux, and have been removed since .net7.

[asiryan]

Thank you very much for your report! You are absolutely right. I think that a little later I will transfer the stack to another core for working with images

[songshizhao]

sys.drawing namespace also can't be used in UWP (not supported), use bytes or stream is better, I think.

[asiryan]

@bnuzhouwei , @songshizhao you can use float[][,] methods

https://github.com/FaceONNX/FaceONNX/blob/9f1fd201d9801dcad20a68f4c67f3545b1faaccc/netstandard/FaceONNX/face/classes/FaceDetector.cs#L91

[Webreaper]

Any chance you could update the examples to use this new method? Then it would be properly cross-platform. I'm going to use ImageSharp to load the bitmap, but it would be great if there was an example showing either ImageSharp or SkiaSharp bitmaps.

[Misaka12456]

Any chance you could update the examples to use this new method? Then it would be properly cross-platform. I'm going to use ImageSharp to load the bitmap, but it would be great if there was an example showing either ImageSharp or SkiaSharp bitmaps.

I asked ChatGPT and it gave me this example:

using SkiaSharp;
using System.Numerics;

public float[][,] ConvertSKBitmapToFloatArray(SKBitmap bitmap)
{
    // Ensure the bitmap is in BGR8888 format
    if (bitmap.ColorType != SKColorType.Rgb && bitmap.AlphaType != SKAlphaType.Opaque)
    {
        bitmap = bitmap.Copy(SKColorType.Rgb, SKAlphaType.Opaque);
    }

    // Lock pixels to get access to pixel data
    using (var info = bitmap.PeekPixels())
    {
        if (info == null)
        {
            throw new InvalidOperationException("Failed to peek pixels from bitmap.");
        }

        var bytes = info.GetPixels();
        var width = bitmap.Width;
        var height = bitmap.Height;

        // Convert byte array to float array
        var floatArray = new float[3, height, width];

        for (int y = 0; y < height; y++)
        {
            for (int x = 0; x < width; x++)
            {
                var index = (y * width + x) * 4; // Each pixel is 4 bytes
                var b = bytes[index];
                var g = bytes[index + 1];
                var r = bytes[index + 2];

                // Normalize values
                floatArray[0, y, x] = b / 255.0f; // Blue
                floatArray[1, y, x] = g / 255.0f; // Green
                floatArray[2, y, x] = r / 255.0f; // Red
            }
        }

        return floatArray;
    }
}

Maybe helpful?

[asiryan]

Any chance you could update the examples to use this new method? Then it would be properly cross-platform. I'm going to use ImageSharp to load the bitmap, but it would be great if there was an example showing either ImageSharp or SkiaSharp bitmaps.

@Webreaper , @Misaka12456 or you can try something like this https://github.com/FaceONNX/FaceONNX/blob/main/netstandard/Examples/FaceEmbeddingsClassification/Program.cs

[Webreaper]

I already submitted a PR (https://github.com/FaceONNX/FaceONNX/pull/22) which fixes this and makes it cross-platform.

[Misaka12456]

I already submitted a PR (#22) which fixes this and makes it cross-platform.

I actually use it in Emgu.CV Mat, but I didn't find a best solution... @asiryan Do you have any other example?

[asiryan]

I already submitted a PR (#22) which fixes this and makes it cross-platform.

I actually use it in Emgu.CV Mat, but I didn't find a best solution... @asiryan Do you have any other example?

Do you need an example with Emgu.CV?

[Misaka12456]

yeah.... planned to use faceonnx for a real-time face detection

[asiryan]

@Misaka12456 I think you can find information about images and their transformations here https://www.emgu.com/wiki/index.php/Working_with_Images

[Misaka12456]

@Misaka12456 I think you can find information about images and their transformations here https://www.emgu.com/wiki/index.php/Working_with_Images

thx