120 likes | 284 Views
Various Methods on Face Recognition. ZJU Xu Hanze xhz1992@gmail.com. Various Methods on:. Dimensionality Reduction LDA( Linear Discriminant Analysis ) PCA( Principal Component Analysis) ... Classifier SRC( Sparse Representation-based Classification) SVM( Support Vector Machine ) ….
E N D
Various Methods on Face Recognition ZJU XuHanze xhz1992@gmail.com
Various Methods on: • Dimensionality Reduction • LDA(Linear Discriminant Analysis) • PCA(Principal Component Analysis) • ... Classifier • SRC(Sparse Representation-based Classification) • SVM(Support Vector Machine) • …
What is Face Recognition? • An application: Given a face picture, find it in a face dataset. • A tool: To prove a certain algorithm of dimensionality reduction & classifier is right. This is why there is a red cross on the initial title. • A game: Full of interest & challenge • …
A Demo Thanks to XieYanan, Wang Yilun, Chen Yusen, Long Qiuyu.
LDA • Supervised Learning • Ronald Fisher, 1936, Fisher’s Linear discriminant • Labeled points Low dimension
LDA max J(w) !
LDA N>2: Lagrange Multiplier: O(N^3)
PCA • Unsupervised Learning • Max Variance(*) • Smallest Loss • Really Dimensionality Reduction
PCA Dim D -> Dim M (M < D) Repeat for M times.
SRC • Sparse Representation-based Classification • Which is SPARSE? 1. A = [A1 A2 .... Ak] (m×n, m<<n) 2. for some vector x that has non-zero components corresponding to the columns of Aj, we have y = Ax 3. x is a sparse vector