Václav Hlaváč

Pose primitive based human action recognition in videos or still images

This paper presents a method for recognizing human actions based on pose primitives. In learning mode, the parameters representing poses and activities are estimated from videos. In run mode, the method can be used both for videos or still images. For recognizing pose primitives, we extend a Histogram of Oriented Gradient (HOG) based descriptor to better cope with articulated poses and cluttered background. Action classes are represented by histograms of poses primitives. For sequences, we incorporate the local temporal context by means of n-gram expressions. Action recognition is based on a simple histogram comparison. Unlike the mainstream video surveillance approaches, the proposed method does not rely on background subtraction or dynamic features and thus allows for action recognition in still images.

Ten lectures on statistical and structural pattern recognition

Preface. Lecture 1. Bayesian statistical decision making. Lecture 2. Non-Bayesian statistical decision making. Lecture 3. Two statistical models of the recognised object. Lecture 4. Learning in pattern recognition. Lecture 5. Linear discriminant function. Lecture 6. Unsupervised Learning. Lecture 7. Mutual relationship of statistical and structural recognition. Lecture 8. Recognition of Markovian sequences. Lecture 9. Regular languages and corresponding pattern recognition tasks. Lecture 10. Context-free languages, their 2-D generalisation, related tasks. Bibliography. Index.

Epipolar Geometry of Panoramic Cameras

This paper presents fundamental theory and design of central panoramic cameras. Panoramic cameras combine a convex hyperbolic or parabolic mirror with a perspective camera to obtain a large field of view. We show how to design a panoramic camera with a tractable geometry and we propose a simple calibration method. We derive the image formation function for such a camera. The main contribution of the paper is the derivation of the epipolar geometry between a pair of panoramic cameras. We show that the mathematical model of a central panoramic camera can be decomposed into two central projections and therefore allows an epipolar geometry formulation. It is shown that epipolar curves are conics and their equations are derived. The theory is tested in experiments with real data.

Multi-class support vector machine

We propose a transformation from the multi-class support vector machine (SVM) classification problem to the single-class SVM problem which is more convenient for optimization. The proposed transformation is based on simplifying the original problem and employing the Kesler construction which can be carried out by the use of properly defined kernel only. The experiments conducted indicate that the proposed method is comparable with the one-against-all decomposition solved by the state-of-the-art sequential minimal optimizer algorithm.

Efficient MRF deformation model for non-rigid image matching

We propose a novel MRF-based model for deformable image matching. Given two images, the task is to estimate a mapping from one image to the other maximizing the quality of the match. We consider mappings defined by a discrete deformation field constrained to preserve 2D continuity. We pose the task as finding MAP configurations of a pairwise MRF. We propose a more compact MRF representation of the problem which leads to a weaker, though computationally more tractable, linear programming relaxation -the approximation technique we choose to apply. The number of dual LP variables grows linearly with the search window side, rather than quadratically as in previous approaches. To solve the relaxed problem (suboptimally), we apply TRW-S (Sequential Tree-Reweighted Message passing) algorithm [13, 5]. Using our representation and the chosen optimization scheme, we are able to match much wider deformations than was considered previously in global optimization framework. We further elaborate on continuity and data terms to achieve more appropriate description of smooth deformations. The performance of our technique is demonstrated on both synthetic and real-world experiments.

Rendering real-world objects using view interpolation

Presents a new approach to rendering arbitrary views of real-world 3D objects of complex shapes. We propose to represent an object by a sparse set of corresponding 2D views, and to construct any other view as a combination of these reference views. We show that this combination can be linear, assuming proximity of the views, and we suggest how the visibility of constructed points can be determined. Our approach makes it possible to avoid difficult 3D reconstruction, assuming only rendering is required. Moreover, almost no calibration of views is needed. We present preliminary results on real objects, indicating that the approach is feasible.<<ETX>>

An iterative algorithm learning the maximal margin classifier

Abstract A simple learning algorithm for maximal margin classifiers (also support vector machines with quadratic cost function) is proposed. We build our iterative algorithm on top of the Schlesinger–Kozinec algorithm (S–K-algorithm) from 1981 which finds a maximal margin hyperplane with a given precision for separable data. We suggest a generalization of the S–K-algorithm (i) to the non-linear case using kernel functions and (ii) for non-separable data. The requirement in memory storage is linear to the data. This property allows the proposed algorithm to be used for large training problems. The resulting algorithm is simple to implement and as the experiments showed competitive to the state-of-the-art algorithms. The implementation of the algorithm in Matlab is available. We tested the algorithm on the problem aiming at recognition poor quality numerals.

Improving stability of feature selection methods

An improper design of feature selection methods can often lead to incorrect conclusions. Moreover, it is not generally realised that functional values of the criterion guiding the search for the best feature set are random variables with some probability distribution. This contribution examines the influence of several estimation techniques on the consistency of the final result. We propose an entropy based measure which can assess the stability of feature selection methods with respect to perturbations in the data. Results show that filters achieve a better stability and performance if more samples are employed for the estimation, i.e., using leave-one-out cross-validation, for instance. However, the best results for wrappers are acquired with the 50/50 holdout validation.

Greedy Algorithm for a Training Set Reduction in the Kernel Methods

We propose a technique for a training set approximation and its usage in kernel methods. The approach aims to represent data in a low dimensional space with possibly minimal representation error which is similar to the Principal Component Analysis (PCA). In contrast to the PCA, the basis vectors of the low dimensional space used for data representation are properly selected vectors from the training set and not as their linear combinations. The basis vectors can be selected by a simple algorithm which has low computational requirements and allows on-line processing of huge data sets. The proposed method was used to approximate training sets of the Support Vector Machines and Kernel Fisher Linear Discriminant which are known method for learning classifiers. The experiments show that the proposed approximation can significantly reduce the complexity of the found classifiers (the number of the support vectors) while retaining their accuracy.

Experience in System Design for Human-Robot Teaming in Urban Search and Rescue

The paper describes experience with applying a user-centric design methodology in developing systems for human-robot teaming in Urban Search & Rescue. A human-robot team consists of several robots (rovers/UGVs, microcopter/UAVs), several humans at an off-site command post (mission commander, UGV operators) and one on-site human (UAV operator). This system has been developed in close cooperation with several rescue organizations, and has been deployed in a real-life tunnel accident use case. The human-robot team jointly explores an accident site, communicating using a multi-modal team interface, and spoken dialogue. The paper describes the development of this complex socio-technical system per se, as well as recent experience in evaluating the performance of this system.