Summary: To appear in Proceedings of IEEE International Conference on Computer Vision (ICCV), October 2007.
ClassMap: Efficient Multiclass Recognition via Embeddings
Vassilis Athitsos 1 , Alexandra Stefan 2 , Quan Yuan 2 , and Stan Sclaroff 2
1 Computer Science and Engineering Department, University of Texas at Arlington, USA
2 Computer Science Department, Boston University, USA
Abstract
In many computer vision applications, such as face
recognition and hand pose estimation, we need systems that
can recognize a very large number of classes. Large margin
classification methods, such as AdaBoost and SVMs, often
provide competitive accuracy rates, but at the cost of eval­
uating a large number of binary classifiers. We propose
an embedding­based method for efficient multiclass recog­
nition. In our method, patterns and classes are mapped to
vectors in such a way that patterns and their associated
classes tend to get mapped close to each other. This way,
given a test pattern, a small set of candidate classes can
be identified efficiently using simple vector comparisons. In
experiments with 3D hand pose recognition (2430 classes)
and face recognition (535 classes), our method is between 3

Source: Athitsos, Vassilis - Department of Computer Science and Engineering, University of Texas at Arlington

Collections: Computer Technologies and Information Sciences