Tentative schedule and topics
DAY 1: FEATURE LEARNING FOR FACE RECOGNITION & EVALUATION
FACE RECOGNITION (WEIHONG DENG)
· Background, Pipeline and Historical Overview
· Recent Progress and Trends
· Basics and Importance of face alignment
FROM SUBSPACE to FEATURE LEARNING-1 (WEIHONG DENG)
· Subspace learning: PCA versus LDA + hands-on
· Basic techniques – revision of basic knowledge
· Basic feature: LBP, HOG, and Gabor + hands-on
· Discussion: Open problems in face recognition techniques
PERFORMANCE EVALUATION (NORMAN POH)
· Evaluating biometric performance with hands-on
· Experimental design of biometric systems
· Confidence estimation + hands-on
INTRA-CLASS VARIATIONS OF BIOMETRIC DATA (MOHAMMAD SABRI)
- Iranian Electronic National ID Card Program
DAY 2: METRIC LEARNING FOR FACE RECOGNITION
FROM SUBSPACE to FEATURE LEARNING-2 (WEIHONG DENG)
· Local feature learning: DFD, CBFD, PCANet
· Hands-on exercise session 1
· Deep feature learning: Deepface and DeepID
· Relationship between Subspace and Feature learning
METRIC LEARNING (WEIHONG DENG)
· Mahalanobis metric learning
· Sparse reconstruction Metric Learning
· Discriminative deep metric learning
· Hands-on exercise session 2
· Discussion: Open problems in learning techniques
DAY 3: 3D FACE RECOGNITION
INTRODUCTION TO 3D FACE RECOGNITION (PAUL KOPPEN)
· 2D photos, stereo matching, 3D spatial coordinates
· Photometric and geometric normalization
· 3D alignment and fitting – ICP, TPS, annotation, degrees of freedom
· Hands-on exercise
· 3D registration, fixed topology
· 3D model construction – matrix representation, PCA
· Hands-on exercise
ADVANCED TOPICS (PAUL KOPPEN)
· 3D Facial analysis – PCA, Non-negative Matrix Factorization
· Automatic annotation – curvature, symmetry, geometric properties, viewpoint invariance
· Open topics and challenges – discussion
DAY 4: AVANCED MACHINE LEARNING FOR FACE RECOGNITION AND FACE LIVENESS
ADVANCED MACHNE LEARNING TECHNIQUES (MANUEL GUNTHER)
· Deep learning for face recognition
· Parts-based face recognition
· Hands-on (using Python or Matlab)
FACIAL ANTISPOOFING (NORMAN POH)
· Facial anti spoofing techniques
· Facial anti spoofing using Dynamic Mode Decomposition + hands-on
CERTIFICATE PRESENTATION CEREMONY