Randoom a Michael Friis production

Webcam face detection in C# using Emgu CV

Some time ago I wrote a post on how to do face detection in C# using OpenCV. I’ve since begun using the Emgu CV wrapper instead of opencvdotnet. Emgu CV is much better, in active development and it even runs on Mono. Two gotchas:

1. You don’t have to install OpenCV, but instead have to copy the relevant dlls (included with the Emgu CV download) to the folder where you code executes.
2. Open CV and X64 are not friends. If you’re running X64 Windows (and unless you are up to recompiling OpenCV) you have to make sure your app is compiled to X86, instead of the usual “Any CPU”.
3. Remember to add PictureBox as per the original tutorial.

Here’s sample code:

using System;
using System.Windows.Forms;
using System.Drawing;
using Emgu.CV;
using Emgu.Util;
using Emgu.CV.Structure;
using Emgu.CV.CvEnum;

namespace opencvtut
{
    public partial class Form1 : Form
    {
		private Capture cap;
		private HaarCascade haar;

        public Form1()
        {
            InitializeComponent();
        }

        private void timer1_Tick(object sender, EventArgs e)
        {
		using (Image nextFrame = cap.QueryFrame())
		{
			if (nextFrame != null)
			{
				// there's only one channel (greyscale), hence the zero index
				//var faces = nextFrame.DetectHaarCascade(haar)[0];
				Image grayframe = nextFrame.Convert();
				var faces =
					grayframe.DetectHaarCascade(
						haar, 1.4, 4,
						HAAR_DETECTION_TYPE.DO_CANNY_PRUNING,
						new Size(nextFrame.Width/8, nextFrame.Height/8)
						)[0];

				foreach (var face in faces)
				{
					nextFrame.Draw(face.rect, new Bgr(0,double.MaxValue,0), 3);
				}
				pictureBox1.Image = nextFrame.ToBitmap();
			}
		}
        }

        private void Form1_Load(object sender, EventArgs e)
        {
            // passing 0 gets zeroth webcam
			cap = new Capture(0);
            // adjust path to find your xml
			haar = new HaarCascade(
                "..\\..\\..\\..\\lib\\haarcascade_frontalface_alt2.xml");
        }
    }
}

[...] UPDATE: Recommend EmguCV for C# wrapping OpenCV, read an updated guide. [...]

Thank you for that code.

Does anyone have some experience about training the HaarClassifier to create xml files for other objects…

Thanks, Sir. A question I want to consult you, how to access my webcam? When I run the code “cap = new Capture(0);” with my wired camera, there is no problem. However, I don’t know how to access my webcam. (I know the address of my webcam)

Have you tried running emgu on a shared hosting server that has trust=Medium forced? it doesn’t seem to like it at all.

@dee Yeah — that would make sense since EmguCV runs on top of OpenCV which is un-managed C++ code.

hello sir
very good your document
tanks

i’m iranian

I used this code to get Facedetection working in C#

however I have this problem when i complie.

It says that ‘var’ could not b found

and ‘Emgu.CV.HarrCascade’ does not contain a definition for ‘rect’

and ‘face’ cannot be declared in this scope becaue it would give a different meaning to ‘face, which is already used in a ‘parent or current’.

I have this problem when i complie:
using (Image nextFrame = cap.QueryFrame())

Cannot convert type ‘Emgu.CV.Image’ to ‘System.Drawing.Image’

Thanks for help. Lada

Thanks for posting your code. This is really helpful. For those getting the compile error, you’ll need to specify params to the Image object to ensure that it’s using Engu.CV.Image and not System.Drawing.Image:

using (Image nextFrame = cap.QueryFrame())
and
Image grayframe = nextFrame.Convert();

Ah, I see that the blog software is stripping out the greater-than and less-than so my code snippets got edited. Let me try again:

using (Image<Bgr, Byte> nextFrame = cap.QueryFrame())

and

Image<Gray, Byte> grayframe = nextFrame.Convert<Gray, Byte>();

Hi. When i run this code i get an exception at the line

Image grayframe = nextFrame.Convert();

and it says that this “conversion is not supported by OpenCV”
Now can anyone tel me how to solve it?

hiii
sir
can u suggest me how o train the databse on your own..
bcoz for side view the results are worst..
or if u could suggest me a way to detect side face of a human in the video with arnd 80 % accuracy..

sonu ->> http://note.sonots.com/SciSoftware/haartraining.html#t1a1f262

Can someone tell me how to extract the face detected and then do some extra detection ? I just need to get the face’s image croped in all frames…

thanks in advance