##Profile Detection Using LBP-Trained Cascades
- Explore other two face recognition algorithms (this will add data to Chapter 5):
[1] EigenFaceRecognizer
Ptr<FaceRecognizer> createEigenFaceRecognizer(int num_components=0, double threshold=DBL_MAX)¶[2] LBPHFaceRecognizer
Ptr<FaceRecognizer> createLBPHFaceRecognizer(int radius=1, int neighbors=8, int grid_x=8, int grid_y=8, double threshold=DBL_MAX) - Save the 'confidence' value of your chosen face in a variable for both frontal and sideview images:
void recognizeProfileFaces(Mat face_scaled, Ptr<FaceRecognizer> modelProfile)
{
modelProfile -> predict(face_scaled, label2, confidence2);
tallyProfile(label2, confidence2);
//~ cout << "LabelSide: " << label2 << endl;
//~ cout << "Confidence: " << confidence2 << endl;
}- After chosing the face, set a threshold for confidence in both frontal and sideview images based on experiments:
//Functions tallyFront and tallyProfile tallies the number of recognized
//faces for a given video sequence or person. The highest number of tallied
//label will be set as maxLabel1 or maxLabel2 and will be considered as
//the predicted label of the system.
void tallyFront(int label1, double confidence1){
if (label1 == 0) a_1++;
if (label1 == 1) b_1++;
if (label1 == 2) c_1++;
if (label1 == 3) d_1++;
if (label1 == 4) e_1++;
if (label1 == 5) f_1++;
if (label1 == 6) g_1++;
if (label1 == 7) h_1++;
if (label1 == 8) i_1++;
if (label1 == 9) j_1++;
//~ cout << a_1 << " " << b_1 << " " << c_1 << " " << d_1 << " "
//~ << e_1 << " " << f_1 << " " << g_1 << " " << h_1 << " "
//~ << i_1 << " " << j_1 << endl;
}Maybe,
if (confidenceChoice1 > frontalThreshold && confidenceChoice2 > sideThreshold)
// Recognition success!- You'll have to perform more experiments and gather more data. (At least 30 different people!)
- Add Screenshots
- Update me with the accuracy