/// <summary>
/// Routine to train face recognizer with sample images
/// </summary>
/*private void TrainRecognizer(string root)
* {
* // This one was actually used to train the recognizer. I didn't push much effort and satisfied once it
* // distinguished all detected faces on the sample image, for the real-world application you might want to
* // refer to the following documentation:
* // OpenCV documentation and samples: http://docs.opencv.org/3.0-beta/modules/face/doc/facerec/tutorial/facerec_video_recognition.html
* // Training sets overview: https://www.kairos.com/blog/60-facial-recognition-databases
* // Another OpenCV doc: http://docs.opencv.org/2.4/modules/contrib/doc/facerec/facerec_tutorial.html#face-database
*
* int id = 0;
* var ids = new List<int>();
* var mats = new List<Mat>();
* var namesList = new List<string>();
*
* foreach (string dir in Directory.GetDirectories(root))
* {
* string name = System.IO.Path.GetFileNameWithoutExtension(dir);
* if (name.StartsWith("-"))
* continue;
*
* namesList.Add(name);
* UnityEngine.Debug.LogFormat("{0} = {1}", id, name);
*
* foreach (string file in Directory.GetFiles(dir))
* {
* var bytes = File.ReadAllBytes(file);
* var texture = new UnityEngine.Texture2D(2, 2);
* texture.LoadImage(bytes); // <--- this one has changed in Unity 2017 API and on that version must be changed
*
* ids.Add(id);
*
* // each loaded texture is converted to OpenCV Mat, turned to grayscale (assuming we have RGB source) and resized
* var mat = Unity.TextureToMat(texture);
* mat = mat.CvtColor(ColorConversionCodes.BGR2GRAY);
* if (requiredSize.Width > 0 && requiredSize.Height > 0)
* mat = mat.Resize(requiredSize);
* mats.Add(mat);
* }
* id++;
* }
*
* names = namesList.ToArray();
*
* // train recognizer and save result for the future re-use, while this isn't quite necessary on small training sets, on a bigger set it should
* // give serious performance boost
* recognizer.Train(mats, ids);
* recognizer.Save(root + "/face-recognizer.xml");
* }*/
#endregion
/// <summary>
/// Initializes scene
/// </summary>
protected virtual void Awake()
{
// classifier
FileStorage storageFaces = new FileStorage(faces.text, FileStorage.Mode.Read | FileStorage.Mode.Memory);
cascadeFaces = new CascadeClassifier();
if (!cascadeFaces.Read(storageFaces.GetFirstTopLevelNode()))
{
throw new System.Exception("FaceProcessor.Initialize: Failed to load faces cascade classifier");
}
// recognizer
// There are three available face recognition algorithms in current version of the OpenCV library (please, refer to the OpenCV documentation for details)
// Our particular training set was trained and saved with FisherFaceRecognizer() and shuld not work with others, however, you can refer to the "TrainRecognizer"
// method defined above to instructions and sample code regarding training your own recognizer from the scratch
//recognizer = FaceRecognizer.CreateLBPHFaceRecognizer();
//recognizer = FaceRecognizer.CreateEigenFaceRecognizer();
recognizer = FaceRecognizer.CreateFisherFaceRecognizer();
// This pre-trained set was quite tiny and contained only those 5 persons that are detected and recognized on the image. We took 5 photos for each person from
// public images on Google, for a real-world application you will need much more sample data for each persona, for more info refer to the OpenCV documentation
// (there are some links in the "TrainRecognizer" sample function
recognizer.Load(new FileStorage(recognizerXml.text, FileStorage.Mode.Read | FileStorage.Mode.Memory));
// label names
names = new string[] { "Cooper", "DeGeneres", "Nyongo", "Pitt", "Roberts", "Spacey" };
}
static void Main(string[] args)
{
var images = new List <ImageInfo>();
var imageId = 0;
foreach (var dir in new DirectoryInfo(@"..\..\Images").GetDirectories())
{
var groupId = int.Parse(dir.Name.Replace("s", string.Empty)) - 1;
foreach (var imageFile in dir.GetFiles("*.pgm"))
{
images.Add(new ImageInfo
{
Image = new Mat(imageFile.FullName, ImreadModes.GrayScale),
ImageId = imageId++,
ImageGroupId = groupId
});
}
}
var model = FaceRecognizer.CreateFisherFaceRecognizer();
model.Train(images.Select(x => x.Image), images.Select(x => x.ImageGroupId));
var rnd = new Random();
var randomImageId = rnd.Next(0, images.Count - 1);
var testSample = images[randomImageId];
Console.WriteLine("Actual group: {0}", testSample.ImageGroupId);
Cv2.ImShow("actual", testSample.Image);
var predictedGroupId = model.Predict(testSample.Image);
Console.WriteLine("Predicted group: {0}", predictedGroupId);
Cv2.WaitKey(0);
}
/// <summary>
///
/// </summary>
/// <param name="numComponents"></param>
/// <param name="threshold"></param>
/// <returns></returns>
public static BasicFaceRecognizer CreateFisherFaceRecognizer(
int numComponents = 0, double threshold = Double.MaxValue)
{
return(FaceRecognizer.CreateFisherFaceRecognizer(numComponents, threshold));
}
/*
* SK: Metóda, ktorá sa zavolá, pre vygenerovanie otázky
* ENG: Method, which is called for generating question
*/
void GenerateQuestion()
{
if (QnA.Count > 0)
{
currentQuestion = Random.Range(0, QnA.Count);
QuestionText.text = QnA[currentQuestion].Questiion;
Image.texture = QnA[currentQuestion].sample;
Mat image = Unity.TextureToMat(QnA[currentQuestion].sample);
// Deteguje tváre
var gray = image.CvtColor(ColorConversionCodes.BGR2GRAY);
Cv2.EqualizeHist(gray, gray);
// deteguje zhodné regióny (Faces bounding)
FileStorage storageFaces = new FileStorage(Faces.text, FileStorage.Mode.Read | FileStorage.Mode.Memory);
cascadeFaces = new CascadeClassifier();
if (!cascadeFaces.Read(storageFaces.GetFirstTopLevelNode()))
{
throw new System.Exception("FaceProcessor.Initialize: Failed to load Faces cascade classifier");
}
recognizer = FaceRecognizer.CreateFisherFaceRecognizer();
recognizer.Load(new FileStorage(RecognizerXml.text, FileStorage.Mode.Read | FileStorage.Mode.Memory));
// popisky (labels)
names = new string[] { "Cooper", "DeGeneres", "Nyongo", "Pitt", "Roberts", "Spacey" };
OpenCvSharp.Rect[] rawFaces = cascadeFaces.DetectMultiScale(gray, 1.1, 6);
foreach (var faceRect in rawFaces)
{
var grayFace = new Mat(gray, faceRect);
if (requiredSize.Width > 0 && requiredSize.Height > 0)
{
grayFace = grayFace.Resize(requiredSize);
}
int label = -1;
/*SK:
* pokús sa rozpoznať tvár:
* confidence" je v princípe vzdialenosť od vzorky k najbližšej známej tvári
* 0 je nejaká „ideálna zhoda“
*/
/*ENG:
* now try to recognize the face:
* confidence" here is actually a misguide. in fact, it's "distance from the sample to the closest known face".
* 0 being some "ideal match"
*/
double confidence = 0.0;
recognizer.Predict(grayFace, out label, out confidence);
faceName = names[label];
int line = 0;
const int textPadding = 2;
const double textScale = 2.0;
string messge = String.Format("{0}", names[label], (int)confidence);
var textSize = Cv2.GetTextSize(messge, HersheyFonts.HersheyPlain, textScale, 1, out line);
var textBox = new OpenCvSharp.Rect(
faceRect.X + (faceRect.Width - textSize.Width) / 2 - textPadding,
faceRect.Bottom,
textSize.Width + textPadding * 2,
textSize.Height + textPadding * 2
);
faceName = names[label];
Debug.Log(faceName);
}
// Priradenie obrázku textúre Image komponentu na scéne
// Asign image to the texture on Image component
var texture = Unity.MatToTexture(image);
var rawImage = Image;
rawImage.texture = texture;
var transform = Image.GetComponent <RectTransform>();
transform.sizeDelta = new Vector2(image.Width, image.Height);
for (int i = 0; i < Options.Length; i++)
{
Options[i].transform.GetChild(0).GetComponent <Text>().text = QnA[currentQuestion].Answers[i];
if (faceName == Options[i].transform.GetChild(0).GetComponent <Text>().text)
{
Options[i].GetComponent <AnswerScript>().isCorrect = true;
}
else
{
Options[i].GetComponent <AnswerScript>().isCorrect = false;
}
}
}
else
{
GameOver();
}
}
public void CreateAndDisposeFisher()
{
var recognizer = FaceRecognizer.CreateFisherFaceRecognizer();
recognizer.Dispose();
}
