Showing papers by "French Institute for Research in Computer Science and Automation published in 2008"

Learning realistic human actions from movies

Abstract: The aim of this paper is to address recognition of natural human actions in diverse and realistic video settings. This challenging but important subject has mostly been ignored in the past due to several problems one of which is the lack of realistic and annotated video datasets. Our first contribution is to address this limitation and to investigate the use of movie scripts for automatic annotation of human actions in videos. We evaluate alternative methods for action retrieval from scripts and show benefits of a text-based classifier. Using the retrieved action samples for visual learning, we next turn to the problem of action classification in video. We present a new method for video classification that builds upon and extends several recent ideas including local space-time features, space-time pyramids and multi-channel non-linear SVMs. The method is shown to improve state-of-the-art results on the standard KTH action dataset by achieving 91.8% accuracy. Given the inherent problem of noisy labels in automatic annotation, we particularly investigate and show high tolerance of our method to annotation errors in the training set. We finally apply the method to learning and classifying challenging action classes in movies and show promising results.

Hamming Embedding and Weak Geometric Consistency for Large Scale Image Search

Abstract: This paper improves recent methods for large scale image search. State-of-the-art methods build on the bag-of-features image representation. We, first, analyze bag-of-features in the framework of approximate nearest neighbor search. This shows the sub-optimality of such a representation for matching descriptors and leads us to derive a more precise representation based on 1) Hamming embedding (HE) and 2) weak geometric consistency constraints (WGC). HE provides binary signatures that refine the matching based on visual words. WGC filters matching descriptors that are not consistent in terms of angle and scale. HE and WGC are integrated within the inverted file and are efficiently exploited for all images, even in the case of very large datasets. Experiments performed on a dataset of one million of images show a significant improvement due to the binary signature and the weak geometric consistency constraints, as well as their efficiency. Estimation of the full geometric transformation, i.e., a re-ranking step on a short list of images, is complementary to our weak geometric consistency constraints and allows to further improve the accuracy.

Discriminative learned dictionaries for local image analysis

Abstract: Sparse signal models have been the focus of much recent research, leading to (or improving upon) state-of-the-art results in signal, image, and video restoration. This article extends this line of research into a novel framework for local image discrimination tasks, proposing an energy formulation with both sparse reconstruction and class discrimination components, jointly optimized during dictionary learning. This approach improves over the state of the art in texture segmentation experiments using the Brodatz database, and it paves the way for a novel scene analysis and recognition framework based on simultaneously learning discriminative and reconstructive dictionaries. Preliminary results in this direction using examples from the Pascal VOC06 and Graz02 datasets are presented as well.

Vehicle Ad Hoc networks: applications and related technical issues

Abstract: This article presents a comprehensive survey of the state-of-the-art for vehicle ad hoc networks. We start by reviewing the possible applications that can be used in VANETs, namely, safety and user applications, and by identifying their requirements. Then, we classify the solutions proposed in the literature according to their location in the open system interconnection reference model and their relationship to safety or user applications. We analyze their advantages and shortcomings and provide our suggestions for a better approach. We also describe the different methods used to simulate and evaluate the proposed solutions. Finally, we conclude with suggestions for a general architecture that can form the basis for a practical VANET.

Computational geometry algorithms library

Abstract: The CGAL C++ library offers geometric data structures and algorithms that are reliable, efficient, easy to use, and easy to integrate in existing software. Use of de facto standard libraries like CGAL increases productivity, because they allow software developers to focus on the application layer. This course is an overview of CGAL geometric algorithms and data structures. The lectures cover:•CGAL for 2D vector graphics, including Boolean operations on Bezier curves, offsets, simplification, and geometry on the sphere.•CGAL for 3D point sets, including principal component analysis, bounding volumes, simplification, outlier removal, normal estimation, normal orientation, denoising, triangulation, and surface reconstruction.•CGAL for mesh-based modeling and processing, including Boolean operations, convex decomposition, simplification, and parameterization.•CGAL for mesh generation, including surface and volume mesh generation, from 3D images, implicit functions, or polyhedral surfaces.The introductory lecture covers non-geometric topics: the exact geometric computing paradigm that makes CGAL reliable without sacrificing efficiency and the generic programming paradigm that facilitates integration into existing software.

Symmetric Tensors and Symmetric Tensor Rank

Abstract: A symmetric tensor is a higher order generalization of a symmetric matrix. In this paper, we study various properties of symmetric tensors in relation to a decomposition into a symmetric sum of outer product of vectors. A rank-1 order-$k$ tensor is the outer product of $k$ nonzero vectors. Any symmetric tensor can be decomposed into a linear combination of rank-1 tensors, each of which is symmetric or not. The rank of a symmetric tensor is the minimal number of rank-1 tensors that is necessary to reconstruct it. The symmetric rank is obtained when the constituting rank-1 tensors are imposed to be themselves symmetric. It is shown that rank and symmetric rank are equal in a number of cases and that they always exist in an algebraically closed field. We will discuss the notion of the generic symmetric rank, which, due to the work of Alexander and Hirschowitz [J. Algebraic Geom., 4 (1995), pp. 201-222], is now known for any values of dimension and order. We will also show that the set of symmetric tensors of symmetric rank at most $r$ is not closed unless $r=1$.

Revisiting the LuGre friction model

Abstract: In this article we first review properties of the LuGre model, including zero-slip displacement, invariance, and passivity. An extension to include velocity-dependent microdamping is also discussed. The resulting model is then used to analyze stick-slip motion. The analysis shows that stick-slip motion modeled by the LuGre model is a stiff system with different behavior in the stick and slip modes as well as dramatic transitions between these modes. The dependence of limit cycles on parameters is discussed along with the notion of rate dependence.

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation

Abstract: Scatterplots remain one of the most popular and widely-used visual representations for multidimensional data due to their simplicity, familiarity and visual clarity, even if they lack some of the flexibility and visual expressiveness of newer multidimensional visualization techniques. This paper presents new interactive methods to explore multidimensional data using scatterplots. This exploration is performed using a matrix of scatterplots that gives an overview of the possible configurations, thumbnails of the scatterplots, and support for interactive navigation in the multidimensional space. Transitions between scatterplots are performed as animated rotations in 3D space, somewhat akin to rolling dice. Users can iteratively build queries using bounding volumes in the dataset, sculpting the query from different viewpoints to become more and more refined. Furthermore, the dimensions in the navigation space can be reordered, manually or automatically, to highlight salient correlations and differences among them. An example scenario presents the interaction techniques supporting smooth and effortless visual exploration of multidimensional datasets.

Bilateral filtering : theory and applications

Abstract: 1: Introduction 2: From Gaussian Convolution to Bilateral Filter 3: Applications 4: Efficient Implementation 5: Relationship between BF and Other Methods or Framework 6: Extensions of Bilateral Filtering 7: Conclusions. Acknowledgements. References.

Clustered Multi-Task Learning: A Convex Formulation

Abstract: In multi-task learning several related tasks are considered simultaneously, with the hope that by an appropriate sharing of information across tasks, each task may benefit from the others. In the context of learning linear functions for supervised classification or regression, this can be achieved by including a priori information about the weight vectors associated with the tasks, and how they are expected to be related to each other. In this paper, we assume that tasks are clustered into groups, which are unknown beforehand, and that tasks within a group have similar weight vectors. We design a new spectral norm that encodes this a priori assumption, without the prior knowledge of the partition of tasks into groups, resulting in a new convex optimization formulation for multi-task learning. We show in simulations on synthetic examples and on the IEDB MHC-I binding dataset, that our approach outperforms well-known convex methods for multi-task learning, as well as related non convex methods dedicated to the same problem.

Symmetric Log-Domain Diffeomorphic Registration: A Demons-Based Approach

Abstract: Modern morphometric studies use non-linear image registration to compare anatomies and perform group analysis. Recently, log-Euclidean approaches have contributed to promote the use of such computational anatomy tools by permitting simple computations of statistics on a rather large class of invertible spatial transformations. In this work, we propose a non-linear registration algorithm perfectly fit for log-Euclidean statistics on diffeomorphisms. Our algorithm works completely in the log-domain, i.e. it uses a stationary velocity field. This implies that we guarantee the invertibility of the deformation and have access to the true inverse transformation. This also means that our output can be directly used for log-Euclidean statistics without relying on the heavy computation of the log of the spatial transformation. As it is often desirable, our algorithm is symmetric with respect to the order of the input images. Furthermore, we use an alternate optimization approach related to Thirion's demons algorithm to provide a fast non-linear registration algorithm. First results show that our algorithm outperforms both the demons algorithm and the recently proposed diffeomorphic demons algorithm in terms of accuracy of the transformation while remaining computationally efficient.

Symmetric tensors and symmetric tensor rank

Abstract: A symmetric tensor is a higher order generalization of a symmetric matrix. In this paper, we study various properties of symmetric tensors in relation to a decomposition into a sum of symmetric outer product of vectors. A rank-1 order-k tensor is the outer product of k non-zero vectors. Any symmetric tensor can be decomposed into a linear combination of rank-1 tensors, each of them being symmetric or not. The rank of a symmetric tensor is the minimal number of rank-1 tensors that is necessary to reconstruct it. The symmetric rank is obtained when the constituting rank-1 tensors are imposed to be themselves symmetric. It is shown that rank and symmetric rank are equal in a number of cases, and that they always exist in an algebraically closed field. We will discuss the notion of the generic symmetric rank, which, due to the work of Alexander and Hirschowitz, is now known for any values of dimension and order. We will also show that the set of symmetric tensors of symmetric rank at most r is not closed, unless r = 1.

Brain-Computer Interfaces, Virtual Reality, and Videogames

Abstract: Major challenges must be tackled for brain-computer interfaces to mature into an established communications medium for VR applications, which will range from basic neuroscience studies to developing optimal peripherals and mental gamepads and more efficient brain-signal processing techniques.

Ten Challenges for Ontology Matching

Abstract: This paper aims at analyzing the key trends and challenges of the ontology matching field. The main motivation behind this work is the fact that despite many component matching solutions that have been developed so far, there is no integrated solution that is a clear success, which is robust enough to be the basis for future development, and which is usable by non expert users. In this paper we first provide the basics of ontology matching with the help of examples. Then, we present general trends of the field and discuss ten challenges for ontology matching, thereby aiming to direct research into the critical path and to facilitate progress of the field.

Quantum hydrodynamic model for the nonlinear electron dynamics in thin metal films

Abstract: A quantum hydrodynamic (fluid) model, derived from the Wigner-Poisson equations, is used to investigate the ultrafast electron dynamics in thin metal films. The hydrodynamic equations, which include exchange and correlation effects, can be combined into a single nonlinear Schr\"odinger-type equation. The fluid model is first benchmarked against a density-functional calculation for the ground state, with good agreement between the two approaches. The ultrafast nonlinear electron dynamics is then investigated and compared to recent semiclassical results obtained with a Vlasov-Poisson approach.

Recent Advances in Remeshing of Surfaces

Abstract: Remeshing is a key component of many geometric algorithms, including modeling, editing, animation and simulation. As such, the rapidly developing field of geometry processing has produced a profusion of new remeshing techniques over the past few years. In this paper we survey recent developments in remeshing of surfaces, focusing mainly on graphics applications. We classify the techniques into five categories based on their end goal: structured, compatible, high quality, feature and error-driven remeshing. We limit our description to the main ideas and intuition behind each technique, and a brief comparison between some of the techniques. We also list some open questions and directions for future research.

OpenAlea: a visual programming and component-based software platform for plant modelling.

Abstract: As illustrated by the approaches presented during the 5th FSPM workshop (Prusinkiewicz and Hanan 2007, and this issue), the development of functional-structural plant models requires an increasing amount of computer modeling. All these models are developed by different teams in various contexts and with different goals. Efficient and flexible computational frameworks are required to augment the interaction between these models, their reusability, and the possibility to compare them on identical datasets. In this paper, we present an open-source platform, OpenAlea, that provides a user-friendly environment for modelers, and advanced deployment methods. OpenAlea allows researchers to build models using a visual programming interface and provides a set of tools and models dedicated to plant modeling. Models and algorithms are embedded in OpenAlea components with well defined input and output interfaces that can be easily interconnected to form more complex models and define more macroscopic components. The system architecture is based on the use of a general purpose, high-level, object-oriented script language, Python, widely used in other scientific areas. We briefly present the rationale that underlies the architectural design of this system and we illustrate the use of the platform to assemble several heterogeneous model components and to rapidly prototype a complex modeling scenario.

Communication-optimal parallel and sequential QR and LU factorizations

Abstract: We present parallel and sequential dense QR factorization algorithms that are both optimal (up to polylogarithmic factors) in the amount of communication they perform, and just as stable as Householder QR. We prove optimality by extending known lower bounds on communication bandwidth for sequential and parallel matrix multiplication to provide latency lower bounds, and show these bounds apply to the LU and QR decompositions. We not only show that our QR algorithms attain these lower bounds (up to polylogarithmic factors), but that existing LAPACK and ScaLAPACK algorithms perform asymptotically more communication. We also point out recent LU algorithms in the literature that attain at least some of these lower bounds.

Proceedings of the 17th ACM conference on Information and knowledge management

Abstract: It is our great pleasure to welcome you to the 17th ACM Conference on Information and Knowledge Management -- CIKM'08. Since 1992, the ACM Conference on Information and Knowledge Management (CIKM) has been successfully bringing together leading researchers and developers from the database, information retrieval, and knowledge management communities. The purpose of the conference is to identify challenging problems facing the development of future knowledge and information systems, and to shape future research directions through the publication of high quality, applied and theoretical research findings. In CIKM 2008, we continued the tradition of promoting collaboration among the general areas of databases, information retrieval, and knowledge management. This year's call for papers attracted almost 800 submissions from Asia, Canada, Europe, Africa, and the United States. The program committee accepted 132 papers and 103 posters giving CIKM'08 an acceptance rate of 17%.

OctoPocus: a dynamic guide for learning gesture-based command sets

Abstract: We describe OctoPocus, an example of a dynamic guide that combines on-screen feedforward and feedback to help users learn, execute and remember gesture sets. OctoPocus can be applied to a wide range of single-stroke gestures and recognition algorithms and helps users progress smoothly from novice to expert performance. We provide an analysis of the design space and describe the results of two experi-ments that show that OctoPocus is significantly faster and improves learning of arbitrary gestures, compared to con-ventional Help menus. It can also be adapted to a mark-based gesture set, significantly improving input time compared to a two-level, four-item Hierarchical Marking menu.

Unifying discriminative visual codebook generation with classifier training for object category recognition

Abstract: The idea of representing images using a bag of visual words is currently popular in object category recognition. Since this representation is typically constructed using unsupervised clustering, the resulting visual words may not capture the desired information. Recent work has explored the construction of discriminative visual codebooks that explicitly consider object category information. However, since the codebook generation process is still disconnected from that of classifier training, the set of resulting visual words, while individually discriminative, may not be those best suited for the classifier. This paper proposes a novel optimization framework that unifies codebook generation with classifier training. In our approach, each image feature is encoded by a sequence of ldquovisual bitsrdquo optimized for each category. An image, which can contain objects from multiple categories, is represented using aggregates of visual bits for each category. Classifiers associated with different categories determine how well a given image corresponds to each category. Based on the performance of these classifiers on the training data, we augment the visual words by generating additional bits. The classifiers are then updated to incorporate the new representation. These two phases are repeated until the desired performance is achieved. Experiments compare our approach to standard clustering-based methods and with state-of-the-art discriminative visual codebook generation. The significant improvements over previous techniques clearly demonstrate the value of unifying representation and classification into a single optimization framework.

A Simple Modification in CMA-ES Achieving Linear Time and Space Complexity

Abstract: This paper proposes a simple modification of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) for high dimensional objective functions, reducing the internal time and space complexity from quadratic to linear. The covariance matrix is constrained to be diagonal and the resulting algorithm, sep-CMA-ES, samples each coordinate independently. Because the model complexity is reduced, the learning rate for the covariance matrix can be increased. Consequently, on essentially separable functions, sep-CMA-ES significantly outperforms CMA-ES . For dimensions larger than a hundred, even on the non-separable Rosenbrock function, the sep-CMA-ES needs fewer function evaluations than CMA-ES .

Alternating Projections on Manifolds

Abstract: We prove that if two smooth manifolds intersect transversally, then the method of alternating projections converges locally at a linear rate. We bound the speed of convergence in terms of the angle between the manifolds, which in turn we relate to the modulus of metric regularity for the intersection problem, a natural measure of conditioning. We discuss a variety of problem classes where the projections are computationally tractable, and we illustrate the method numerically on a problem of finding a low-rank solution of a matrix equation.

Discriminative Sparse Image Models for Class-Specific Edge Detection and Image Interpretation

Abstract: Sparse signal models learned from data are widely used in audio, image, and video restoration. They have recently been generalized to discriminative image understanding tasks such as texture segmentation and feature selection. This paper extends this line of research by proposing a multiscale method to minimize least-squares reconstruction errors and discriminative cost functions under ?0 or ?1 regularization constraints. It is applied to edge detection, category-based edge selection and image classification tasks. Experiments on the Berkeley edge detection benchmark and the PASCAL VOC'05 and VOC'07 datasets demonstrate the computational efficiency of our algorithm and its ability to learn local image descriptions that effectively support demanding computer vision tasks.

Viewpoint-independent object class detection using 3D Feature Maps

Abstract: This paper presents a 3D approach to multi-view object class detection. Most existing approaches recognize object classes for a particular viewpoint or combine classifiers for a few discrete views. We propose instead to build 3D representations of object classes which allow to handle viewpoint changes and intra-class variability. Our approach extracts a set of pose and class discriminant features from synthetic 3D object models using a filtering procedure, evaluates their suitability for matching to real image data and represents them by their appearance and 3D position. We term these representations 3D Feature Maps. For recognizing an object class in an image we match the synthetic descriptors to the real ones in a 3D voting scheme. Geometric coherence is reinforced by means of a robust pose estimation which yields a 3D bounding box in addition to the 2D localization. The precision of the 3D pose estimation is evaluated on a set of images of a calibrated scene. The 2D localization is evaluated on the PASCAL 2006 dataset for motorbikes and cars, showing that its performance can compete with state-of-the-art 2D object detectors.

Cross-View Action Recognition from Temporal Self-similarities

Abstract: This paper concerns recognition of human actions under view changes. We explore self-similarities of action sequences over time and observe the striking stability of such measures across views. Building upon this key observation we develop an action descriptor that captures the structure of temporal similarities and dissimilarities within an action sequence. Despite this descriptor not being strictly view-invariant, we provide intuition and experimental validation demonstrating the high stability of self-similarities under view changes. Self-similarity descriptors are also shown stable under action variations within a class as well as discriminative for action recognition. Interestingly, self-similarities computed from different image features possess similar properties and can be used in a complementary fashion. Our method is simple and requires neither structure recovery nor multi-view correspondence estimation. Instead, it relies on weak geometric properties and combines them with machine learning for efficient cross-view action recognition. The method is validated on three public datasets, it has similar or superior performance compared to related methods and it performs well even in extreme conditions such as when recognizing actions from top views while using side views for training only.