// Start training from the collected faces.
// The face recognition algorithm can be one of these and perhaps more, depending on your version of OpenCV, which must be atleast v2.4.1:
// "FaceRecognizer.Eigenfaces": Eigenfaces, also referred to as PCA (Turk and Pentland, 1991).
// "FaceRecognizer.Fisherfaces": Fisherfaces, also referred to as LDA (Belhumeur et al, 1997).
// "FaceRecognizer.LBPH": Local Binary Pattern Histograms (Ahonen et al, 2006).
public static BasicFaceRecognizer LearnCollectedFaces(List <Mat> preprocessedFaces, List <int> faceLabels, string facerecAlgorithm = "FaceRecognizer.Eigenfaces")
{
BasicFaceRecognizer model = null;
Debug.Log("Learning the collected faces using the [" + facerecAlgorithm + "] algorithm ...");
if (facerecAlgorithm == "FaceRecognizer.Fisherfaces")
{
model = FisherFaceRecognizer.create();
}
else if (facerecAlgorithm == "FaceRecognizer.Eigenfaces")
{
model = EigenFaceRecognizer.create();
}
if (model == null)
{
Debug.LogError("ERROR: The FaceRecognizer algorithm [" + facerecAlgorithm + "] is not available in your version of OpenCV. Please update to OpenCV v2.4.1 or newer.");
//exit(1);
}
// Do the actual training from the collected faces. Might take several seconds or minutes depending on input!
MatOfInt labels = new MatOfInt();
labels.fromList(faceLabels);
model.train(preprocessedFaces, labels);
return(model);
}
private void Run()
{
List <Mat> images = new List <Mat> ();
List <int> labelsList = new List <int> ();
MatOfInt labels = new MatOfInt();
images.Add(Imgcodecs.imread(image_0_filepath, 0));
images.Add(Imgcodecs.imread(image_1_filepath, 0));
labelsList.Add(0);
labelsList.Add(1);
labels.fromList(labelsList);
Mat testSampleMat = Imgcodecs.imread(sample_image_filepath, 0);
int testSampleLabel = 0;
// foreach (Mat item in images) {
// Debug.Log ("images.ToString " + item.ToString ());
// }
// foreach (int item in labelsList) {
// Debug.Log ("labels.ToString " + item.ToString ());
// }
int[] predictedLabel = new int[1];
double[] predictedConfidence = new double[1];
BasicFaceRecognizer faceRecognizer = EigenFaceRecognizer.create();
faceRecognizer.train(images, labels);
faceRecognizer.predict(testSampleMat, predictedLabel, predictedConfidence);
Debug.Log("Predicted class: " + predictedLabel [0] + " / " + "Actual class: " + testSampleLabel);
Debug.Log("Confidence: " + predictedConfidence [0]);
Mat predictedMat = images [predictedLabel [0]];
Mat baseMat = new Mat(testSampleMat.rows(), predictedMat.cols() + testSampleMat.cols(), CvType.CV_8UC1);
predictedMat.copyTo(baseMat.submat(new OpenCVForUnity.CoreModule.Rect(0, 0, predictedMat.cols(), predictedMat.rows())));
testSampleMat.copyTo(baseMat.submat(new OpenCVForUnity.CoreModule.Rect(predictedMat.cols(), 0, testSampleMat.cols(), testSampleMat.rows())));
Imgproc.putText(baseMat, "Predicted", new Point(10, 15), Imgproc.FONT_HERSHEY_SIMPLEX, 0.4, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, "Confidence:", new Point(5, 25), Imgproc.FONT_HERSHEY_SIMPLEX, 0.2, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, " " + predictedConfidence [0], new Point(5, 33), Imgproc.FONT_HERSHEY_SIMPLEX, 0.2, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, "TestSample", new Point(predictedMat.cols() + 10, 15), Imgproc.FONT_HERSHEY_SIMPLEX, 0.4, new Scalar(255), 1, Imgproc.LINE_AA, false);
Texture2D texture = new Texture2D(baseMat.cols(), baseMat.rows(), TextureFormat.RGBA32, false);
Utils.matToTexture2D(baseMat, texture);
gameObject.GetComponent <Renderer> ().material.mainTexture = texture;
}
/// <summary>
/// Raises the load button event.
/// </summary>
public void OnLoadButton()
{
Debug.Log("User clicked [Load] button.");
// Restart everything!
dispose();
if (facerecAlgorithm == "FaceRecognizer.Fisherfaces")
{
model = Face.createFisherFaceRecognizer();
}
else if (facerecAlgorithm == "FaceRecognizer.Eigenfaces")
{
model = Face.createEigenFaceRecognizer();
}
if (model == null)
{
Debug.LogError("ERROR: The FaceRecognizer algorithm [" + facerecAlgorithm + "] is not available in your version of OpenCV. Please update to OpenCV v2.4.1 or newer.");
m_mode = MODES.MODE_DETECTION;
return;
}
// load the train data.
model.load(Application.temporaryCachePath + "/traindata.yml");
int maxLabel = (int)Core.minMaxLoc(model.getLabels()).maxVal;
if (maxLabel <= 0)
{
Debug.Log("load failure.");
model.Dispose();
model = null;
m_mode = MODES.MODE_DETECTION;
return;
}
// Restore the save data.
m_numPersons = maxLabel + 1;
for (int i = 0; i < m_numPersons; ++i)
{
m_latestFaces.Add(i);
preprocessedFaces.Add(Imgcodecs.imread(Application.temporaryCachePath + "/preprocessedface" + i + ".jpg", 0));
if (preprocessedFaces [i].empty())
{
preprocessedFaces [i] = new Mat(faceHeight, faceWidth, CvType.CV_8UC1, new Scalar(128));
}
faceLabels.Add(i);
}
// go to the recognition mode!
m_mode = MODES.MODE_RECOGNITION;
}
private void Run()
{
List <Mat> images = new List <Mat>();
List <int> labelsList = new List <int>();
MatOfInt labels = new MatOfInt();
images.Add(Imgcodecs.imread(facerec_0_bmp_filepath, 0));
images.Add(Imgcodecs.imread(facerec_1_bmp_filepath, 0));
labelsList.Add(0); //积极
labelsList.Add(1); //消极
labels.fromList(labelsList);
Mat testSampleMat = Imgcodecs.imread(facerec_sample_bmp_filepath, 0);
int testSampleLabel = 0;
int[] predictedLabel = new int[1];
double[] predictedConfidence = new double[1];
BasicFaceRecognizer faceRecognizer = EigenFaceRecognizer.create();
//faceRecognizer.train(images, labels); //会清空之前模型
faceRecognizer.update(images, labels); //不会清空load的模型
faceRecognizer.predict(testSampleMat, predictedLabel, predictedConfidence);
//faceRecognizer.save(Application.dataPath + "/Cascades/train_face.txt"); //生成一个txt
Debug.Log("Predicted class: " + predictedLabel[0] + " / " + "Actual class: " + testSampleLabel);
Debug.Log("Confidence: " + predictedConfidence[0]);
Mat predictedMat = images[predictedLabel[0]];
Mat baseMat = new Mat(testSampleMat.rows(), predictedMat.cols() + testSampleMat.cols(), CvType.CV_8UC1);
predictedMat.copyTo(baseMat.submat(new OpenCVForUnity.Rect(0, 0, predictedMat.cols(), predictedMat.rows())));
testSampleMat.copyTo(baseMat.submat(new OpenCVForUnity.Rect(predictedMat.cols(), 0, testSampleMat.cols(), testSampleMat.rows())));
Imgproc.putText(baseMat, "Predicted", new Point(10, 15), Core.FONT_HERSHEY_SIMPLEX, 0.4, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, "Confidence:", new Point(5, 25), Core.FONT_HERSHEY_SIMPLEX, 0.2, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, " " + predictedConfidence[0], new Point(5, 33), Core.FONT_HERSHEY_SIMPLEX, 0.2, new Scalar(255), 1, Imgproc.LINE_AA, false);
Imgproc.putText(baseMat, "TestSample", new Point(predictedMat.cols() + 10, 15), Core.FONT_HERSHEY_SIMPLEX, 0.4, new Scalar(255), 1, Imgproc.LINE_AA, false);
Texture2D t2d = new Texture2D(baseMat.width(), baseMat.height());
Utils.matToTexture2D(baseMat, t2d);
Sprite sp = Sprite.Create(t2d, new UnityEngine.Rect(0, 0, t2d.width, t2d.height), Vector2.zero);
m_srcImage.sprite = sp;
m_srcImage.preserveAspect = true;
}
// Disposal to restart all.
private void dispose()
{
m_selectedPerson = -1;
m_numPersons = 0;
m_latestFaces.Clear();
faceLabels.Clear();
prev_identity = -1;
prev_similarity = 100000000.0d;
foreach (Mat face in preprocessedFaces)
{
if (face != null && !face.IsDisposed)
{
face.Dispose();
}
}
preprocessedFaces.Clear();
if (old_prepreprocessedFace != null && !old_prepreprocessedFace.IsDisposed)
{
old_prepreprocessedFace.Dispose();
}
old_prepreprocessedFace = null;
if (prev_prepreprocessedFace != null && !prev_prepreprocessedFace.IsDisposed)
{
prev_prepreprocessedFace.Dispose();
}
prev_prepreprocessedFace = null;
if (reconstructedFace != null && !reconstructedFace.IsDisposed)
{
reconstructedFace.Dispose();
}
reconstructedFace = null;
if (model != null && !model.IsDisposed)
{
model.Dispose();
model = null;
}
}
// Generate an approximately reconstructed face by back-projecting the eigenvectors & eigenvalues of the given (preprocessed) face.
public static Mat ReconstructFace(BasicFaceRecognizer model, Mat preprocessedFace)
{
// Since we can only reconstruct the face for some types of FaceRecognizer models (ie: Eigenfaces or Fisherfaces),
// we should surround the OpenCV calls by a try/catch block so we don't crash for other models.
try
{
// Get some required data from the FaceRecognizer model.
Mat eigenvectors = model.getEigenVectors();
Mat averageFaceRow = model.getMean();
int faceHeight = preprocessedFace.rows();
// Project the input image onto the PCA subspace.
Mat projection = subspaceProject(eigenvectors, averageFaceRow, preprocessedFace.reshape(1, 1));
//printMatInfo(projection, "projection");
// Generate the reconstructed face back from the PCA subspace.
Mat reconstructionRow = subspaceReconstruct(eigenvectors, averageFaceRow, projection);
//printMatInfo(reconstructionRow, "reconstructionRow");
// Convert the float row matrix to a regular 8-bit image. Note that we
// shouldn't use "getImageFrom1DFloatMat()" because we don't want to normalize
// the data since it is already at the perfect scale.
// Make it a rectangular shaped image instead of a single row.
Mat reconstructionMat = reconstructionRow.reshape(1, faceHeight);
// Convert the floating-point pixels to regular 8-bit uchar pixels.
Mat reconstructedFace = new Mat(reconstructionMat.size(), CvType.CV_8UC1);
reconstructionMat.convertTo(reconstructedFace, CvType.CV_8UC1, 1, 0);
//printMatInfo(reconstructedFace, "reconstructedFace");
return(reconstructedFace);
}
catch (CvException e)
{
Debug.Log("WARNING: Missing FaceRecognizer properties." + e);
return(new Mat());
}
}
// Show the internal face recognition data, to help debugging.
public static void ShowTrainingDebugData(BasicFaceRecognizer model, int faceWidth, int faceHeight)
{
// TODO...
}
// Main loop that runs forever, until the user hits Escape to quit.
private void recognizeAndTrainUsingWebcam(Mat cameraFrame, CascadeClassifier faceCascade, CascadeClassifier eyeCascade1, CascadeClassifier eyeCascade2)
{
if (cameraFrame != null && cameraFrame.empty())
{
Debug.LogError("ERROR: Couldn't grab the next camera frame.");
}
// Get a copy of the camera frame that we can draw onto.
Mat displayedFrame = cameraFrame;
int cx;
float current_processingTime = Time.realtimeSinceStartup;
float processingTimeDiff_seconds = (current_processingTime - old_processingTime);
if (processingTimeDiff_seconds > CHANGE_IN_SECONDS_FOR_PROCESSING)
{
// Run the face recognition system on the camera image. It will draw some things onto the given image, so make sure it is not read-only memory!
int identity = -1;
// Find a face and preprocess it to have a standard size and contrast & brightness.
Rect faceRect = new Rect(); // Position of detected face.
Rect searchedLeftEye = new Rect(), searchedRightEye = new Rect(); // top-left and top-right regions of the face, where eyes were searched.
Point leftEye = new Point(), rightEye = new Point(); // Position of the detected eyes.
Mat preprocessedFace = PreprocessFace.GetPreprocessedFace(displayedFrame, faceWidth, faceCascade, eyeCascade1, eyeCascade2, preprocessLeftAndRightSeparately, ref faceRect, ref leftEye, ref rightEye, ref searchedLeftEye, ref searchedRightEye);
bool gotFaceAndEyes = false;
if (preprocessedFace != null && !preprocessedFace.empty())
{
gotFaceAndEyes = true;
}
// Draw an anti-aliased rectangle around the detected face.
if (faceRect.width > 0)
{
Imgproc.rectangle(displayedFrame, faceRect.tl(), faceRect.br(), YELLOW, 2, Imgproc.LINE_AA, 0);
// Draw light-blue anti-aliased circles for the 2 eyes.
Scalar eyeColor = LIGHT_BLUE;
if (leftEye.x >= 0) // Check if the eye was detected
{
Imgproc.circle(displayedFrame, new Point(faceRect.x + leftEye.x, faceRect.y + leftEye.y), 6, eyeColor, 1, Imgproc.LINE_AA, 0);
}
if (rightEye.x >= 0) // Check if the eye was detected
{
Imgproc.circle(displayedFrame, new Point(faceRect.x + rightEye.x, faceRect.y + rightEye.y), 6, eyeColor, 1, Imgproc.LINE_AA, 0);
}
}
prev_prepreprocessedFace = preprocessedFace;
if (m_mode == MODES.MODE_DETECTION)
{
// Don't do anything special.
}
else if (m_mode == MODES.MODE_COLLECT_FACES)
{
// Check if we have detected a face.
if (gotFaceAndEyes)
{
// Check if this face looks somewhat different from the previously collected face.
double imageDiff = 10000000000.0d;
if (old_prepreprocessedFace != null && !old_prepreprocessedFace.empty())
{
imageDiff = Recognition.GetSimilarity(preprocessedFace, old_prepreprocessedFace);
}
// Also record when it happened.
double current_time = Time.realtimeSinceStartup;
double timeDiff_seconds = (current_time - old_time);
// Only process the face if it is noticeably different from the previous frame and there has been noticeable time gap.
if ((imageDiff > CHANGE_IN_IMAGE_FOR_COLLECTION) && (timeDiff_seconds > CHANGE_IN_SECONDS_FOR_COLLECTION))
{
// Also add the mirror image to the training set, so we have more training data, as well as to deal with faces looking to the left or right.
Mat mirroredFace = new Mat();
Core.flip(preprocessedFace, mirroredFace, 1);
// Add the face images to the list of detected faces.
preprocessedFaces.Add(preprocessedFace);
preprocessedFaces.Add(mirroredFace);
faceLabels.Add(m_selectedPerson);
faceLabels.Add(m_selectedPerson);
// Keep a reference to the latest face of each person.
m_latestFaces [m_selectedPerson] = preprocessedFaces.Count - 2; // Point to the non-mirrored face.
// Show the number of collected faces. But since we also store mirrored faces, just show how many the user thinks they stored.
Debug.Log("Saved face " + (preprocessedFaces.Count / 2) + " for person " + m_selectedPerson);
// Make a white flash on the face, so the user knows a photo has been taken.
using (Mat displayedFaceRegion = new Mat(displayedFrame, faceRect)) {
Core.add(displayedFaceRegion, DARK_GRAY, displayedFaceRegion);
}
// Keep a copy of the processed face, to compare on next iteration.
old_prepreprocessedFace = preprocessedFace;
old_time = current_time;
}
}
}
else if (m_mode == MODES.MODE_TRAINING)
{
// Check if there is enough data to train from. For Eigenfaces, we can learn just one person if we want, but for Fisherfaces,
// we need atleast 2 people otherwise it will crash!
bool haveEnoughData = true;
if (facerecAlgorithm == "FaceRecognizer.Fisherfaces")
{
if ((m_numPersons < 2) || (m_numPersons == 2 && m_latestFaces [1] < 0))
{
Debug.Log("Warning: Fisherfaces needs atleast 2 people, otherwise there is nothing to differentiate! Collect more data ...");
haveEnoughData = false;
}
}
if (m_numPersons < 1 || preprocessedFaces.Count <= 0 || preprocessedFaces.Count != faceLabels.Count)
{
Debug.Log("Warning: Need some training data before it can be learnt! Collect more data ...");
haveEnoughData = false;
}
if (haveEnoughData)
{
// Start training from the collected faces using Eigenfaces or a similar algorithm.
model = Recognition.LearnCollectedFaces(preprocessedFaces, faceLabels, facerecAlgorithm);
// Show the internal face recognition data, to help debugging.
//if (m_debug)
//Recognition.ShowTrainingDebugData(model, faceWidth, faceHeight);
// Now that training is over, we can start recognizing!
m_mode = MODES.MODE_RECOGNITION;
}
else
{
// Since there isn't enough training data, go back to the face collection mode!
m_mode = MODES.MODE_COLLECT_FACES;
}
}
else if (m_mode == MODES.MODE_RECOGNITION)
{
prev_identity = -1;
prev_similarity = 100000000.0d;
if (reconstructedFace != null && !reconstructedFace.IsDisposed)
{
reconstructedFace.Dispose();
}
reconstructedFace = null;
if (gotFaceAndEyes && (preprocessedFaces.Count > 0) && (preprocessedFaces.Count == faceLabels.Count))
{
// Generate a face approximation by back-projecting the eigenvectors & eigenvalues.
reconstructedFace = Recognition.ReconstructFace(model, preprocessedFace);
// Verify whether the reconstructed face looks like the preprocessed face, otherwise it is probably an unknown person.
double similarity = Recognition.GetSimilarity(preprocessedFace, reconstructedFace);
double confidence = 0.0d;
string outputStr;
if (similarity < UNKNOWN_PERSON_THRESHOLD)
{
int[] predictedLabel = new int [1];
double[] predictedConfidence = new double [1];
// Identify who the person is in the preprocessed face image.
model.predict(preprocessedFace, predictedLabel, predictedConfidence);
identity = predictedLabel [0];
confidence = predictedConfidence [0];
outputStr = identity.ToString();
prev_identity = identity;
}
else
{
// Since the confidence is low, assume it is an unknown person.
outputStr = "Unknown";
}
prev_similarity = similarity;
Debug.Log("Identity: " + outputStr + ". Similarity: " + similarity + ". Confidence: " + confidence);
}
}
else if (m_mode == MODES.MODE_DELETE_ALL)
{
// Restart everything!
dispose();
// Restart in Detection mode.
m_mode = MODES.MODE_DETECTION;
}
else
{
Debug.LogError("ERROR: Invalid run mode " + m_mode);
//exit(1);
}
old_processingTime = current_processingTime;
}
// Show the help, while also showing the number of collected faces. Since we also collect mirrored faces, we should just
// tell the user how many faces they think we saved (ignoring the mirrored faces), hence divide by 2.
strBuilder.Length = 0;
Rect rcHelp = new Rect();
if (m_mode == MODES.MODE_DETECTION)
{
strBuilder.Append("Click [Add Person] when ready to collect faces.");
}
else if (m_mode == MODES.MODE_COLLECT_FACES)
{
strBuilder.Append("Click anywhere to train from your ");
strBuilder.Append(preprocessedFaces.Count / 2);
strBuilder.Append(" faces of ");
strBuilder.Append(m_numPersons);
strBuilder.Append(" people.");
}
else if (m_mode == MODES.MODE_TRAINING)
{
strBuilder.Append("Please wait while your ");
strBuilder.Append(preprocessedFaces.Count / 2);
strBuilder.Append(" faces of ");
strBuilder.Append(m_numPersons);
strBuilder.Append(" people builds.");
}
else if (m_mode == MODES.MODE_RECOGNITION)
{
strBuilder.Append("Click people on the right to add more faces to them, or [Add Person] for someone new.");
}
if (strBuilder.Length > 0)
{
// Draw it with a black background and then again with a white foreground.
// Since BORDER may be 0 and we need a negative position, subtract 2 from the border so it is always negative.
float txtSize = 0.4f;
drawString(displayedFrame, strBuilder.ToString(), new Point(BORDER, -BORDER - 2), BLACK, txtSize); // Black shadow.
rcHelp = drawString(displayedFrame, strBuilder.ToString(), new Point(BORDER + 1, -BORDER - 1), WHITE, txtSize); // White text.
}
// Show the current mode.
strBuilder.Length = 0;
if (m_mode >= 0 && m_mode < MODES.MODE_END)
{
strBuilder.Append(" people builds.");
strBuilder.Append(MODE_NAMES [(int)m_mode]);
drawString(displayedFrame, strBuilder.ToString(), new Point(BORDER, -BORDER - 2 - rcHelp.height), BLACK); // Black shadow
drawString(displayedFrame, strBuilder.ToString(), new Point(BORDER + 1, -BORDER - 1 - rcHelp.height), GREEN); // Green text
}
// Show the current preprocessed face in the top-center of the display.
cx = (displayedFrame.cols() - faceWidth) / 2;
if (prev_prepreprocessedFace != null && !prev_prepreprocessedFace.empty())
{
// Get a RGBA version of the face, since the output is RGBA color.
using (Mat srcRGBA = new Mat(prev_prepreprocessedFace.size(), CvType.CV_8UC4)) {
Imgproc.cvtColor(prev_prepreprocessedFace, srcRGBA, Imgproc.COLOR_GRAY2RGBA);
// Get the destination ROI (and make sure it is within the image!).
Rect dstRC = new Rect(cx, BORDER, faceWidth, faceHeight);
using (Mat dstROI = new Mat(displayedFrame, dstRC)) {
// Copy the pixels from src to dst.
srcRGBA.copyTo(dstROI);
}
}
}
// Draw an anti-aliased border around the face, even if it is not shown.
Imgproc.rectangle(displayedFrame, new Point(cx - 1, BORDER - 1), new Point(cx - 1 + faceWidth + 2, BORDER - 1 + faceHeight + 2), LIGHT_GRAY, 1, Imgproc.LINE_AA, 0);
// Show the most recent face for each of the collected people, on the right side of the display.
m_gui_faces_left = displayedFrame.cols() - BORDER - faceWidth;
m_gui_faces_top = BORDER;
for (int i = 0; i < m_numPersons; i++)
{
int index = m_latestFaces [i];
if (index >= 0 && index < preprocessedFaces.Count)
{
Mat srcGray = preprocessedFaces [index];
if (srcGray != null && !srcGray.empty())
{
// Get a RGBA version of the face, since the output is RGBA color.
using (Mat srcRGBA = new Mat(srcGray.size(), CvType.CV_8UC4)) {
Imgproc.cvtColor(srcGray, srcRGBA, Imgproc.COLOR_GRAY2RGBA);
// Get the destination ROI (and make sure it is within the image!).
int y = Mathf.Min(m_gui_faces_top + i * faceHeight, displayedFrame.rows() - faceHeight);
Rect dstRC = new Rect(m_gui_faces_left, y, faceWidth, faceHeight);
using (Mat dstROI = new Mat(displayedFrame, dstRC)) {
// Copy the pixels from src to dst.
srcRGBA.copyTo(dstROI);
}
}
}
}
}
// Highlight the person being collected, using a red rectangle around their face.
if (m_mode == MODES.MODE_COLLECT_FACES)
{
if (m_selectedPerson >= 0 && m_selectedPerson < m_numPersons)
{
int y = Mathf.Min(m_gui_faces_top + m_selectedPerson * faceHeight, displayedFrame.rows() - faceHeight);
Rect rc = new Rect(m_gui_faces_left, y, faceWidth, faceHeight);
Imgproc.rectangle(displayedFrame, rc.tl(), rc.br(), RED, 3, Imgproc.LINE_AA, 0);
}
}
// Highlight the person that has been recognized, using a green rectangle around their face.
if (m_mode == MODES.MODE_RECOGNITION && prev_identity >= 0 && prev_identity < 1000)
{
int y = Mathf.Min(m_gui_faces_top + prev_identity * faceHeight, displayedFrame.rows() - faceHeight);
Rect rc = new Rect(m_gui_faces_left, y, faceWidth, faceHeight);
Imgproc.rectangle(displayedFrame, rc.tl(), rc.br(), GREEN, 3, Imgproc.LINE_AA, 0);
}
if (m_mode == MODES.MODE_RECOGNITION)
{
if (m_debug)
{
if (reconstructedFace != null && !reconstructedFace.empty())
{
cx = (displayedFrame.cols() - faceWidth) / 2;
Point rfDebugBottomRight = new Point(cx + faceWidth * 2 + 5, BORDER + faceHeight);
Point rfDebugTopLeft = new Point(cx + faceWidth + 5, BORDER);
Rect rfDebugRC = new Rect(rfDebugTopLeft, rfDebugBottomRight);
using (Mat srcRGBA = new Mat(reconstructedFace.size(), CvType.CV_8UC4)) {
Imgproc.cvtColor(reconstructedFace, srcRGBA, Imgproc.COLOR_GRAY2RGBA);
using (Mat dstROI = new Mat(displayedFrame, rfDebugRC)) {
srcRGBA.copyTo(dstROI);
}
}
Imgproc.rectangle(displayedFrame, rfDebugTopLeft, rfDebugBottomRight, LIGHT_GRAY, 1, Imgproc.LINE_AA, 0);
}
}
// Show the confidence rating for the recognition in the mid-top of the display.
cx = (displayedFrame.cols() - faceWidth) / 2;
Point ptBottomRight = new Point(cx - 5, BORDER + faceHeight);
Point ptTopLeft = new Point(cx - 15, BORDER);
// Draw a gray line showing the threshold for an "unknown" person.
Point ptThreshold = new Point(ptTopLeft.x, ptBottomRight.y - (1.0 - UNKNOWN_PERSON_THRESHOLD) * faceHeight);
Imgproc.rectangle(displayedFrame, ptThreshold, new Point(ptBottomRight.x, ptThreshold.y), LIGHT_GRAY, 1, Imgproc.LINE_AA, 0);
// Crop the confidence rating between 0.0 to 1.0, to show in the bar.
double confidenceRatio = 1.0d - Math.Min(Math.Max(prev_similarity, 0.0d), 1.0d);
Point ptConfidence = new Point(ptTopLeft.x, ptBottomRight.y - confidenceRatio * faceHeight);
// Show the light-blue confidence bar.
Imgproc.rectangle(displayedFrame, ptConfidence, ptBottomRight, LIGHT_BLUE, Core.FILLED, Imgproc.LINE_AA, 0);
// Show the gray border of the bar.
Imgproc.rectangle(displayedFrame, ptTopLeft, ptBottomRight, LIGHT_GRAY, 1, Imgproc.LINE_AA, 0);
}
/*
* // If the user wants all the debug data, show it to them!
* if (m_debug)
* {
* Mat face = new Mat();
* if (faceRect.width > 0)
* {
* face = new Mat(cameraFrame, faceRect);
* if (searchedLeftEye.width > 0 && searchedRightEye.width > 0)
* {
* Mat topLeftOfFace = new Mat(face, searchedLeftEye);
* Mat topRightOfFace = new Mat(face, searchedRightEye);
* //imshow("topLeftOfFace", topLeftOfFace);
* //imshow("topRightOfFace", topRightOfFace);
* }
* }
*
* //if (model != null)
* //showTrainingDebugData(model, faceWidth, faceHeight);
* }
*/
}
/// <summary>
/// Raises the load button click event.
/// </summary>
public void LoadModel()
{
string loadDirectoryPath = Path.Combine(Application.persistentDataPath, saveDirectoryName);
if (!Directory.Exists(loadDirectoryPath))
{
Debug.Log("load failure. saved train data file does not exist.");
return;
}
// Restart everything!
dispose();
if (facerecAlgorithm == "FaceRecognizer.Fisherfaces")
{
model = FisherFaceRecognizer.create();
}
else if (facerecAlgorithm == "FaceRecognizer.Eigenfaces")
{
model = EigenFaceRecognizer.create();
}
if (model == null)
{
Debug.LogError("ERROR: The FaceRecognizer algorithm [" + facerecAlgorithm + "] is not available in your version of OpenCV. Please update to OpenCV v2.4.1 or newer.");
m_mode = R_MODES.MODE_DETECTION;
return;
}
// load the train data.
model.read(Path.Combine(loadDirectoryPath, "traindata.yml"));
int maxLabel = (int)Core.minMaxLoc(model.getLabels()).maxVal;
if (maxLabel < 0)
{
Debug.Log("load failure.");
model.Dispose();
return;
}
// Restore the save data.
#if UNITY_WEBGL && !UNITY_EDITOR
string format = "jpg";
#else
string format = "png";
#endif
m_numPersons = maxLabel + 1;
personsNames = new string[m_numPersons];
for (int i = 0; i < m_numPersons; ++i)
{
personsNames[i] = GameManager.instance.personsNames[i];
m_latestFaces.Add(i);
preprocessedFaces.Add(Imgcodecs.imread(Path.Combine(loadDirectoryPath, "preprocessedface" + i + "." + format), 0));
if (preprocessedFaces[i].total() == 0)
{
preprocessedFaces[i] = new Mat(faceHeight, faceWidth, CvType.CV_8UC1, new Scalar(128));
}
faceLabels.Add(i);
}
// go to the recognition mode!
m_mode = R_MODES.MODE_RECOGNITION;
}
