Media IC System Lab Face Recognition A Distributed

Media IC & System Lab Face Recognition: A Distributed Approach Io. T / Video Sensor/ CV / Distributed System Cheng-Yi, Chuang (莊成毅), b 99

Outline • About the topic • Face recognition • Future Object 2

Media IC & System Lab Image/Video/IC Io. T Sensor networking Camera (video sensor) CV/Networking/Comm. /Control CV Anomaly Detection/Face Recognition A distributed computational approach 1. 2. 3. 4. Big data Limited bandwidth Limited computation Limited power 3

two meanings of “distributed” 4

Distributed system framework Filtering Gate way Filtering 5

Face Recognition Block Diagram Recognize what? http: //www. uurmi. com/property/frs. html 6

Principle Component Analysis (PCA) (1/3) • Face dataset Eigenface PCA [U, S, V] = svd(Sigma);

PCA (2/3) • How to choose k? • Projection onto Eigenfaces Ureduce = U(: , 1: k); Z = Ureduce’ * X; • Reconstruction / Classification / Recognition / (Whatever…) X = Ureduce * Z

PCA (3/3) 32 x 32 in grayscale = 1024 dimensions 100 principle components = 100 dimensions (display first 36) PCA Original faces Principle components Recovered faces 9 Stanford ML class by Andrew Ng

Related Techniques Face Detection Face Recognition Haar Features. . . Principle Component Analysis (PCA) Linear Discriminant Analysis (LDA) Local Binary Pattern (LBP) Gabor features. . .

Future Object • Get familiar with Open. CV/MATLAB • Study related techniques & papers • Implement the algorithm • Modify it into distributed manner • Hardware? • Of course, keep thinking, and thinking…… 11

Thanks you! 12