Learning a Dynamic Classification Method to Detect Faces and Identify Facial Expression
- Ramana Isukapalli
- Ahmed Elgammal, Rutgers University
- Russ Greiner, Dept of Computing Science; PI of AICML
A lot of current research in face detection is based on using a 'cascade' of boosted classifiers, after the seminal work of Viola and Jones. Their work uses 'Adaboost', a well known boosting algorithm that separates 'clusters' of positive examples from negatives. This can be used to detect most of the common objects, not just faces. We extend the cascade detection methods by 'learning' a decision tree of boosted classifiers that can discover sub-clusters in any class of objects, where each sub-cluster signifies images that are similar to each other in some way. A second contribution of our work is that it presents a 'dynamic detection' technique to detect objects and also associates each detected object to one of the sub-clusters. Unlike the cascade classification methods, that are static detection methods, our detection algorithm applies the most effective classifier at every stage based on the outcome of the classifiers already applied. To clarify and prove the concepts, we first show how our method discovers sub-clusters in a domain to detect simple geometric shapes and also on another domain to detect cars from the rear view. We then present several examples in the domain of face detection to illustrate how method discovers sub-clusters having similar facial expressions. We also evaluate our work extensively and show that our detection method performs better that the Viola-Jones method on the MIT-CMU database to detect faces. A salient feature of our work is that it can be easily extended to do multiple object detection using binary classifiers.
Citation
R. Isukapalli, A. Elgammal, R. Greiner. "Learning a Dynamic Classification Method to Detect Faces and Identify Facial Expression". IEEE International Workshop on Analysis and Modeling of Faces and Gestures, Springer, pp 70-84, October 2005.| Keywords: | efficient, vision, face recognition, machine learning |
| Category: | In Workshop |
| Web Links: | dblp |
BibTeX
@misc{Isukapalli+al:AMFG200505,
author = {Ramana Isukapalli and Ahmed Elgammal and Russ Greiner},
title = {Learning a Dynamic Classification Method to Detect Faces and
Identify Facial Expression},
Booktitle = {Analysis and Modelling of Faces and Gestures, Second
International Workshop ... Proceedings},
Publisher = "Springer",
Pages = {70-84},
booktitle = {IEEE International Workshop on Analysis and Modeling of Faces
and Gestures},
year = 2005,
}Last Updated: November 20, 2019Submitted by Sabina P
