Selected Readings in Vision and Graphics
edited by Luc Van Gool, Gábor Székely, Markus Gross, Bernt Schiele

Volume 55

Philipp Zehnder

Efficient Multi-Class Object Detection

2009. XX, 252 pages. EUR 64,00. ISBN-10: 3-86628-289-3

ISBN-13: 978-3-86628-289-6

Multi-class object detection plays an important role in semantic image analysis. Usually, the task is solved by running a separate detector for each class, which quickly gets expensive if the number of classes increases. In this thesis, we propose a method to improve the efficiency of multi-class detection by learning shared features between classes. First, two decision tree approaches are investigated. Then, a better solution is developed, the so-called 'shared cascade'. Experiments on real-world data show a significant efficiency improvement compared to standard approches. Furthermore, we propose a method to add a class to an existing shared cascade without training everything from scratch again.

About the author:

Philipp Zehnder studied Information Technology and Electrical Engineering at the ETH Zürich where he graduated as Dipl. El.-Ing. ETH in 2002. Subsequently, he joined the Computer Vision Laboratory at the ETH Zürich where he worked as a research assistant and PhD student. In 2009, he was awarded a PhD degree (Dr. sc. ETH Zürich) for his work about efficient multi-class object detection.

Keywords / Schlagwörter:

Object Detection, Multi-Class, Feature Sharing, Machine Learning, SVM, AdaBoost, Haar Features, Cascade of Classifiers

Reihe " Selected Readings in Vision and Graphics " im Hartung-Gorre Verlag

Direkt bestellen bei / to order directly from: Hartung.Gorre@t-online.de

Hartung-Gorre Verlag / D-78465 Konstanz / Germany

Telefon: +49 (0) 7533 97227 Telefax: +49 (0) 7533 97228

http://www.hartung-gorre.de
eMail: verlag@hartung-gorre.de