Fabrizio Smeraldi

Comparison of face verification results on the XM2VTFS database

Presents results of the face verification contest that was organized in conjunction with International Conference on Pattern Recognition 2000. Participants had to use identical data sets from a large, publicly available multimodal database XM2VTSDB. Training and evaluation was carried out according to an a priori known protocol. Verification results of all tested algorithms have been collected and made public on the XM2VTSDB website, facilitating large scale experiments on classifier combination and fusion. Tested methods included, among others, representatives of the most common approaches to face verification -elastic graph matching, Fishers linear discriminant and support vector machines.

Cognitive navigation based on nonuniform Gabor space sampling, unsupervised growing networks, and reinforcement learning

We study spatial learning and navigation for autonomous agents. A state space representation is constructed by unsupervised Hebbian learning during exploration. As a result of learning, a representation of the continuous two-dimensional (2-D) manifold in the high-dimensional input space is found. The representation consists of a population of localized overlapping place fields covering the 2-D space densely and uniformly. This space coding is comparable to the representation provided by hippocampal place cells in rats. Place fields are learned by extracting spatio-temporal properties of the environment from sensory inputs. The visual scene is modeled using the responses of modified Gabor filters placed at the nodes of a sparse Log-polar graph. Visual sensory aliasing is eliminated by taking into account self-motion signals via path integration. This solves the hidden state problem and provides a suitable representation for applying reinforcement learning in continuous space for action selection. A temporal-difference prediction scheme is used to learn sensorimotor mappings to perform goal-oriented navigation. Population vector coding is employed to interpret ensemble neural activity. The model is validated on a mobile Khepera miniature robot.

Saccadic search with Gabor features applied to eye detection and real-time head tracking

The Gabor decomposition is a ubiquitous tool in computer vision. Nevertheless, it is generally considered computationally demanding for active vision applications. We suggest an attention-driven approach to feature detection inspired by the human saccadic system. A dramatic speedup is achieved by computing the Gabor decomposition only on the points of a sparse retinotopic grid. An off-line eye detection application and a real-time head localisation and tracking system are presented. The real-time system features a novel eyeball-mounted camera designed to simulate the dynamic performance of the human eye and is, to the best of our knowledge, the first example of active vision system based on the Gabor decomposition.

Non-rigid structure from motion using ranklet-based tracking and non-linear optimization

In this paper, we address the problem of estimating the 3D structure and motion of a deformable object given a set of image features tracked automatically throughout a video sequence. Our contributions are twofold: firstly, we propose a new approach to improve motion and structure estimates using a non-linear optimization scheme and secondly we propose a tracking algorithm based on ranklets, a recently developed family of orientation selective rank features. It has been shown that if the 3D deformations of an object can be modeled as a linear combination of shape bases then both its motion and shape may be recovered using an extension of Tomasi and Kanades factorization algorithm for affine cameras. Crucially, these new factorization methods are model free and work purely from video in an unconstrained case: a single uncalibrated camera viewing an arbitrary 3D surface which is moving and articulating. The main drawback of existing methods is that they do not provide correct structure and motion estimates: the motion matrix has a repetitive structure which is not respected by the factorization algorithm. In this paper, we present a non-linear optimization method to refine the motion and shape estimates which minimizes the image reprojection error and imposes the correct structure onto the motion matrix by choosing an appropriate parameterization. Factorization algorithms require as input a set of feature tracks or correspondences found throughout the image sequence. The challenge here is to track the features while the object is deforming and the appearance of the image therefore changing. We propose a model free tracking algorithm based on ranklets, a multi-scale family of rank features that present an orientation selectivity pattern similar to Haar wavelets. A vector of ranklets is used to encode an appearance based description of a neighborhood of each tracked point. Robustness is enhanced by adapting, for each point, the shape of the filters to the structure of the particular neighborhood. A stack of models is maintained for each tracked point in order to manage large appearance variations with limited drift. Our experiments on sequences of a human subject performing different facial expressions show that this tracker provides a good set of feature correspondences for the non-rigid 3D reconstruction algorithm.

Development and automation of a test of impulse control in zebrafish

Deficits in impulse control (difficulties in inhibition of a pre-potent response) are fundamental to a number of psychiatric disorders, but the molecular and cellular basis is poorly understood. Zebrafish offer a very useful model for exploring these mechanisms, but there is currently a lack of validated procedures for measuring impulsivity in fish. In mammals, impulsivity can be measured by examining rates of anticipatory responding in the 5-choice serial reaction time task (5-CSRTT), a continuous performance task where the subject is reinforced upon accurate detection of a briefly presented light in one of five distinct spatial locations. This paper describes the development of a fully-integrated automated system for testing impulsivity in adult zebrafish. We outline the development of our image analysis software and its integration with National Instruments drivers and actuators to produce the system. We also describe an initial validation of the system through a one-generation screen of chemically mutagenized zebrafish, where the testing parameters were optimized.

Non-Rigid Structure from Motion using non-Parametric Tracking and Non-Linear Optimization

In this paper we address the problem of estimating the 3D structure and motion of a deformable non-rigid object from a sequence of uncalibrated images. It has been recently shown that if the deformation is modelled as a linear combination of basis shapes both the motion and the 3D structure of the object may be recovered using an extension of Tomasi and Kanades factorization algorithm for affine cameras. The main drawback of the existing methods is that the non-rigid factorization algorithm does not provide a correct estimate of the motion: the motion matrix has a repetitive structure which is not respected by the factorization algorithm. This also affects the estimation of the 3D shape. In this paper we present a non-linear optimization method which minimizes image reprojection error and imposes the correct structure onto the motion matrix by choosing an appropriate parameterization. In addition, we propose a novel non-rigid tracking algorithm based on the use of ranklets, a multiscale family of rank features. Finally, we show that improved motion and shape estimates are obtained on a real image sequence of a persons face which is moving and changing expression.

Combining Colour and Orientation for Adaptive Particle Filter-based Tracking.

We propose an accurate tracking algorithm based on a multi-feature statistical model. The model combines in a single particle filter colour and gradient-based orientation information. A reliability measure derived from the particle distribution is used to adaptively weigh the contribution of the two features. Furthermore, information from the tracker is used to set the dimension of the filters for the computation of the gradient, effectively solving the scale selection problem. Experiments over a set of real-world sequences show that the adaptive use of colour and orientation information improves over either feature taken separately, both in terms of tracking accuracy and of reduction of lost tracks. Also, the automatic scale selection for the derivative filters results in increased robustness.

Multi-Modal Person Authentication

This paper deals with the elements of a multi-modal person authentication systems. Test procedures for evaluating machine experts as well as machine supervisors based on leave-one-out principle are described. Two independent machine experts on person authentication are presented along with their individual performances. These experts consisted of a face (Gabor features) and a speaker (LPC features) authentication algorithm trained on the M2VTS multi-media database. The expert opinions are combined yielding far better performances by using a trained supervisor based on Bayesian statistics than individual modalities aggregated by averaging.

Facial feature detection by saccadic exploration of the Gabor decomposition

The Gabor decomposition is a ubiquitous tool in computer vision. Nevertheless, it is generally considered computationally demanding for active vision applications. We suggest an attention-driven approach to feature detection inspired by the human saccadic system. A dramatic speedup is achieved by computing the Gabor decomposition only on the points of a sparse retinotopic grid. An application to eye detection is presented. Also, a real-time head detection and tracking system based on our approach is briefly discussed. The system features a novel eyeball-mounted camera designed to mimic the dynamic performance of the human eye and is, to the best of our knowledge, the first example of active vision system based on the Gabor decomposition.

Game Theory Meets Information Security Management

This work addresses the challenge “how do we make better security decisions?” and it develops techniques to support human decision making and algorithms which enable well-founded cyber security decisions to be made. In this paper we propose a game theoretic model which optimally allocates cyber security resources such as administrators’ time across different tasks. We first model the interactions between an omnipresent attacker and a team of system administrators seen as the defender, and we have derived the mixed Nash Equilibria (NE) in such games. We have formulated general-sum games that represent our cyber security environment, and we have proven that the defender’s Nash strategy is also minimax. This result guarantees that independently from the attacker’s strategy the defender’s solution is optimal. We also propose Singular Value Decomposition (SVD) as an efficient technique to compute approximate equilibria in our games. By implementing and evaluating a minimax solver with SVD, we have thoroughly investigated the improvement that Nash defense introduces compared to other strategies chosen by common sense decision algorithms. Our key finding is that a particular NE, which we call weighted NE, provides the most effective defense strategy. In order to validate this model we have used real-life statistics from Hackmageddon, the Verizon 2013 Data Breach Investigation report, and the Ponemon report of 2011. We finally compare the game theoretic defense method with a method which implements a stochastic optimization algorithm.