Pattern Recognition Note :
#1. Supervised / Unsupervised
Supervised -> Classification ( Need to set LABEL )
Unsupervised -> Clustering
class Face{ public: int label = 1; }; class Non_Face{ public: int label = -1; };
#2. Classifiers :
1. ) Bayes : Probability and Statistics
2. ) SVM ( Support Vector Machine ) :
Optimization ( like Genetic Algorithms ), high performance, high complexity.
3. ) NN ( Neural Network )
4. ) Adaboost ( Adaptive Boosting ) :
Supervised classifier, Fast ( performance ≒ SVM ), useful in real-time system.
Assembling classifiers - Combine many low accuracy classifiers (weak learners)
to create a high-accurancy classifier ( strong learners )
5. ) KNN ( k-Nearest Neighbor ) :
Weak learners, Voting system, Simple, but low performance.
Diffcault to implement, and easy to affected by noise.
#3. Procedure :
Input Training Process : ( 1 ) Data collection
↓ / ( 2 ) Feature extraction
Classifiers ( 3 ) Training classifier
↓ \
Output Test Process
#4. Feature Vector :
1. ) Gray value ( useless )
2. ) LBP ( Local Binary Pattern ) - Human Face
3. ) SIFT ( Scale-Invariant Feature Transform ) - Object Recognition
HOG ( Histograms of Orinted Gradients ) /
4. ) Color ( HSI, YIQ, YCbCr ... etc. )