Results 1 
5 of
5
Learning of variability for invariant statistical pattern recognition
 In ECML 2001, 12th European Conference on Machine Learning
, 2001
"... ..."
(Show Context)
A Probabilistic View on Tangent Distance
 In 22nd Symposium of the German Association for Pattern Recognition
, 2000
"... In this paper we present a new probabilistic interpretation of tangent distance, which proved to be very effective in modeling image transformations in object recognition. Descriptions of the resulting distributions in pattern space are given for different possible models of variation, leading to a ..."
Abstract

Cited by 8 (5 self)
 Add to MetaCart
(Show Context)
In this paper we present a new probabilistic interpretation of tangent distance, which proved to be very effective in modeling image transformations in object recognition. Descriptions of the resulting distributions in pattern space are given for different possible models of variation, leading to a natural derivation of tangent distance. Furthermore, a possible generalization is presented and experimental results on the well known US Postal Service database are presented. Invariance of classification algorithms with respect to certain transformations plays an important role in pattern recognition. For example, in recognition of image objects like handwritten digits, invariance with respect to (small) affine variations is desired. One method which can achieve such invariance by using first order approximation of the manifolds generated by the considered transformations is known as tangent distance (TD). It was introduced by Simard et al. [14, 13] and successfully used for pat...
Mixtures of Latent Variable Models for Density Estimation and Classification
, 2000
"... This paper deals with the problem of probability density estimation with the goal of finding a good probabilistic representation of the data. One of the most popular density estimation methods is the Gaussian mixture model (GMM). A promising alternative to GMMS are the recently proposed mixtures of ..."
Abstract

Cited by 2 (0 self)
 Add to MetaCart
This paper deals with the problem of probability density estimation with the goal of finding a good probabilistic representation of the data. One of the most popular density estimation methods is the Gaussian mixture model (GMM). A promising alternative to GMMS are the recently proposed mixtures of latent variable models. Examples of the latter are principal component analysis and factor analysis. The advantage of these models is that they are capable of representing the covariance structure with less parameters by choosing the dimension of a subspace in a suitable way. An empirical evaluation on a large number of data sets shows that mixtures of latent variable models almost always outperform various GMMS both in density estimation and Bayes classifiers. To avoid having to choose a value for the dimension of the latent subspace by a computationally expensive search technique such as crossvalidation, a Bayesian treatment of mixtures of latent variable models is proposed. This framework makes it possible to determine the appropriate dimension during training and experiments illustrate its viability.
Video skimming and summarization based on principal component analysis
 Proceedings of the IFIP/IEEE International Conference on Management of Multimedia Networks and Services
, 2001
"... Abstract. An increasing number of applications such as contentbased multimedia retrieval in a distributed system and lowbitrate video communications, require the efficient processing and transmission of video information. In contentbased video retrieval, video segmentation produces video shots ch ..."
Abstract

Cited by 2 (1 self)
 Add to MetaCart
(Show Context)
Abstract. An increasing number of applications such as contentbased multimedia retrieval in a distributed system and lowbitrate video communications, require the efficient processing and transmission of video information. In contentbased video retrieval, video segmentation produces video shots characterized by a certain degree of visual cohesiveness. The number of relevant video shots returned by the system can be very large, thereby requiring significant transmission bandwidth. In this paper, we present a new algorithm for the representation of visual information contained in video segments. The approach is based on Principal Component Analysis and takes advantage of the characteristics of the data in video shots, and the optimal energy compaction properties of the transform. The algorithm can use additional information about video sequences provided by a video analysis and retrieval system, such as a visual change estimator, and a video object tracking module. 1