4EyesFace-Realtime face detection, tracking, alignment and recognition

1
4EyesFace-Realtime face detection, tracking,
alignment and recognition

• Changbo Hu, Rogerio Feris and Matthew Turk

2
Overview

• Introduction
• Face Detection and Pose tracking
• Face Alignment
• Face Recognition
• Conclusions

3
Introduction
Detection
Pose tracking
Alignment
Recognition
4
Introduction

• Why this is a difficult problem?
• Facial Expressions, Illumination Changes, Pose,
  etc.
• Object
• Develop a fully automatic system, suitable for
  real-time applications to locate and track human
  faces, then to align and recognize the face.
• Evaluate it on a large dataset.

5
Face Detection

• Viola and Jones, 2001
• Simple features, which can be computed very
  fast.
• A variant of Adaboost is used both to select
  the features and to train the classifier.
• Classifiers are combined in a cascade which
  allows background regions of the image to be
  quickly discarded.

6
Face detection
7
Pose tracking
Based on Kentaro Toyamas IFA framework
8
Face Alignment

• Active Appearance Model (AAM)

Statistical Shape Model (PCA)
Statistical Texture Model (PCA)
9
Face alignment

• Problem Partial Occlusion
• Active Wavelet Networks (AWN) (on BMVC03)
• Main idea Replace AAM texture model by a
  wavelet network

10
Face Alignment
Similar performance to AAM in images under normal
conditions.
More robust against partial occlusions.
11
Face Alignment
Using 9 wavelets, the system requires only
3 ms per iteration. In general, at most 10
iterations are sufficiently for good convergence
(PIV 1.6Ghz).
12
Multi-View Face Alignment

• View selection by pose tracker

13
Multi-View Face Alignment
14
Face recognition

• online recognition
• HMM based face recognition

15
Face recognition

• Large dataset evaluation
• FERET DataSet
• 1196 different individuals
• With ground truth of eye corners

16
Face recognition
17
Face recognition
18
Face Recognition
19
Face Recognition
20
Conclusion

• We develop a system to do human face detection,
  tracking, alignment and recognition
• In this system, we invented new methods AWN and
  extent to multi-view AWN
• We implement the related detection and pose
  tracking
• Evaluate our method on large dataset