Multi-View Face Detection and Tracking
                 
1. Learning discriminant features for multi-view face detection
2. Collaborative multi-view face tracking and online learning
3. Small multi-View face detection
Publication:

1. Peng Wang and Qiang Ji, "Learning Discriminant Features for Multi-View Face and Eye Detection",  IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), San Diego, CA, June 2005

2. Peng Wang, Qiang Ji, "Multi-View Face Tracking with Factorial and Switching HMM",  Workshop on Application of Computer Vision (WACV), 2005

3. Peng Wang, Qiang Ji, "Multi-View Face Detection under Complex Scene based on Combined SVMs", International Conference on Pattern Recognition (ICPR), 2004

A Paper in submission to conference

A Paper in submission to journal

1. Learning discriminant features for multi-view face detection
We present  a discriminant analysis method to extract powerful feature to detect multi-view face. The discriminant features are learned from training data to minimize the misclassification
errors, and are combined through AdaBoost to form a face classifier.
                                    Multi-view face detection in video
2. Collaborative Multi-View Face Tracking and Online Learning

To address the drifting problems associated with the traditional object tracking under significant changes in both objects and
environmental conditions, a novel collaborative tracking algorithm is presented. The collaborative tracking method probabilistically
combines measurements from specific face models with measurements from a generic face model in a dynamic Bayesian network for robust
multi-view face tracking. In addition, the probabilistic tracking results are used to incrementally adapt the specific face models to
individuals. The collaborative tracking method can handle large face pose changes, and can efficiently build specific face models online.
          Pose angle estimation during tracking
Demo: face tracking and pose estimation in video
Online learning face appearance model using PPCA model

The each view of a multi-view face is modeled with a Probabilistic PCA (PPCA) model.  The model 

parameters are online incrementally updated using tracking result.  The following figure  shows tracking 

results, and online learned PPCA model for each view.  The columns show mean,  and principal 

components in PPCA models. 
3. Small multi-View face detection under complex scenes based on combined SVM

Given the identified image regions corresponding to the detected human, we then propose
to use appearance-based face detection method. Two specific methods are developed.
We use combined support vector machine (SVM)  to detect the most likely face location in
each region.