Converting Mat to byte[] to send to nvidia triton server

[SCraenen]

Summary of your issue

I want to convert OpenCvSharp Mat object to a byte[] that maintains it's size. To clarify what I mean is if I'd have an image of width and height 640, I want to receive a byte[] of size 640 640 3 4 (w h c 4bytesInFloat).

To clarify the purpose: I will write this byte[] in to system shared memory and transfer it to nvidia triton server to do model inference.

Environment

I'm working on windows, linux x64 and linux arm64.

Example code:

In python the equivalent code would look something like this:

image = cv2.imread(image_path, cv2.IMREAD_COLOR)
image = image.transpose((2, 0, 1))
image = image.astype(dtype=np.float32) / 255
image_bytes = image.tobytes()

Desired result:

I want to implement the same functionality as above with OpenCvSharp.

---- edit ---- it appears that the following code does the trick. I am confused because the bytes I manually inspected did not match my expectation. I created a 3 by 3 image with only full r g b white or black pixels however there were values like 4 in my byte array. It turns out that when I save the image with a bit depth of 8 I get the correct bytes while the original image with bit depth 24 gives invalid colours when loaded with Cv2.ImRead();

Marshal.Copy(image.Data, imageBytes, 0, (int)imageByteCount);

I still need to have the functionality of the transpose which I can't seem to figure out with OpenCvSharp

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.