In the medical field, the use of sensors and wearable devices is now widely regarded as a routine support technique for clinical evaluations. Given the advancements in activity recognition from wearable devices in recent years, it is reasonable to explore the potential of similar data as clinical assessment tools for monitoring the progression of chronic diseases. In the current state of the art, datasets collected using wearable devices for subjects with diseases are relatively rare, and their availability becomes even more limited when focusing on pediatric subjects. Therefore, we decided to record a new dataset in collaboration with the Istituto Giannina Gaslini (Genoa, Italy), a center of excellence in pediatric rheumatology. In this article, we present and describe a dataset collected using accelerometers in FDA-approved wearable devices, positioned on the wrist and ankle of the subjects. This dataset included patients between the ages of 2 and 18 years with chronic diseases and age-matched healthy children as a control group. The activities of daily living to be recorded were selected in collaboration with medical specialists, as the diseases considered may potentially affect functional abilities. This shared dataset could enable the scientific community to develop machine learning-based methods to assess severity and monitor the progression of these diseases in a remote, objective, and non-intrusive manner. The recorded dataset is published and fully accessible on the Harvard Dataverse web portal.

Magnetic Resonance urography (MRU) can be used to evaluate abnormalities of the urinary tract in children, with the advantage of being a non-invasive technique and allowing both morphologic and functional assessments. Today, MRU analysis is usually performed using semi-automatic software that typically requires manual segmentation of kidney and pelvis and other time-consuming interactions. In this work, we propose a deep learning approach to automatize the functional MRU analysis. Our pipeline first employs an Attention U-Net for kidney and pelvis segmentation on morphological magnetic resonance and then an image registration process to align the segmentations on the functional MR. The automatic segmentation of morphological MR has been tested on 107 patients using cross validation, achieving a dice score of$\mathbf{0.87}\pm \mathbf{0.15}$and$\mathbf{0.91}\pm \mathbf{0.11}$for left and right kidney, and a dice score of$\mathbf{0.75}\pm \mathbf{0.24}$and$\mathbf{0.71}\pm \mathbf{0.25}$for the left and right pelvis respectively. These segmentations are used to extract morphological and functional parameters to assess the urinary-tract function in children undergoing analysis. The proposed approach has been integrated into a commercial web viewer (DicomVision 0.18.3) so that it can be used easily by clinical staff. Our tests demonstrate that this automated tool allows for rapid and comprehensive analysis of children MRU, thus laying the pave for its wider exploitation in clinical routines.

Reproducibility of proposed approaches is a crucial element in scientific fields, in order to let other researchers trust published works. Moreover, in order to let authors compare the effectiveness of a novel method to the state of the art, benchmark datasets should be commonly used. Concentrating on the task of activity recognition using data coming from wearable devices with inertial sensors, we have analyzed the reproducibility of proposed approaches with a focus on used datasets. In this work, with a literature review, we have measured what percentage of works in the literature verified their approach using public datasets or sharing the ones created on purpose. At the same time, we have also examined the characteristics of considered datasets, with attention to the amount of data recorded, involved population, and studied activities. Starting from 1289 works retrieved on Scopus, we analyzed in detail 146 of them and found out that approximately one out of three (~33%) used public datasets and that less than one out of three (~28%) of the specially made datasets were shared with the public. Moreover, considering all the examined datasets, 13% of them had restricted access (e.g. requiring requests to authors or subscriptions to websites for a fee) or were offline.

The increased availability of multivariate time-series asks for the development of suitable methods able to holistically analyse them. To this aim, we propose a novel flexible method for data-mining, forecasting and causal patterns detection that leverages the coupling of Hidden Markov Models and Gaussian Graphical Models. Given a multivariate non-stationary time-series, the proposed method simultaneously clusters time points while understanding probabilistic relationships among variables. The clustering divides the time points into stationary sub-groups whose underlying distribution can be inferred through a graphical model. Such coupling can be further exploited to build a time-varying regression model which allows to both make predictions and obtain insights on the presence of causal patterns. We extensively validate the proposed approach on synthetic data showing that it has better performance than the state of the art on clustering, graphical models inference and prediction. Finally, to demonstrate the applicability of our approach in real-world scenarios, we exploit its characteristics to build a profitable investment portfolio. Results show that we are able to improve the state of the art, by going from a -%20 profit to a noticeable 80%.

Automated recognition of Activities of daily living is a significant task since it allows monitoring patients remotely using wearable devices. It could also help doctors and specialists to easily track the health status of elderly people and of patients suffering from a variety of geriatric diseases. Most of the literature regarding activity recognition does not face the problem of classifying activities of interest performed among activities that are not, i.e., unknown activities for the classifier. In this paper, we propose a novel method, based on the acceleration data recorded by wearable smartwatches, aimed at improving the activity recognition accuracy in presence of unknown activities (i.e., in real-world settings). The approach is based on an ensemble of different classifiers, a filtering procedure, and a final voting mechanism. The empirical evaluation, carried out on 8 different subjects, shows that our approach improves the overall F1 score by 5.5%, increases the recognition of unknown activities by 11.1%, and decreases the amount of data wrongly classified as an unknown activity by 15.5%. The first two observed differences are respectively statistically significant (Wilcoxon test p-value<0.01).

In many applications of finance, biology and sociology, complex systems involve entities interacting with each other. These processes have the peculiarity of evolving over time and of comprising latent factors, which influence the system without being explicitly measured. In this work we present latent variable time-varying graphical lasso (LTGL), a method for multivariate time-series graphical modelling that considers the influence of hidden or unmeasurable factors. The estimation of the contribution of the latent factors is embedded in the model which produces both sparse and low-rank components for each time point. In particular, the first component represents the connectivity structure of observable variables of the system, while the second represents the influence of hidden factors, assumed to be few with respect to the observed variables. Our model includes temporal consistency on both components, providing an accurate evolutionary pattern of the system. We derive a tractable optimisation algorithm based on alternating direction method of multipliers, and develop a scalable and efficient implementation which exploits proximity operators in closed form. LTGL is extensively validated on synthetic data, achieving optimal performance in terms of accuracy, structure learning and scalability with respect to ground truth and state-of-the-art methods for graphical inference. We conclude with the application of LTGL to real case studies, from biology and finance, to illustrate how our method can be successfully employed to gain insights on multivariate time-series data.

In this paper, we propose a sparse coding approach to background modeling. The obtained model is based on dictionaries which we learn and keep up to date as new data are provided by a video camera. We observe that, without dynamic events, video frames may be seen as noisy data belonging to the background. Over time, such background is subject to local and global changes due to variable illumination conditions, camera jitter, stable scene changes, and intermittent motion of background objects. To capture the locality of some changes, we propose a space-variant analysis where we learn a dictionary of atoms for each image patch, the size of which depends on the background variability. At run time, each patch is represented by a linear combination of the atoms learnt online. A change is detected when the atoms are not sufficient to provide an appropriate representation, and stable changes over time trigger an update of the current dictionary. Even if the overall procedure is carried out at a coarse level, a pixel-wise segmentation can be obtained by comparing the atoms with the patch corresponding to the dynamic event. Experiments on benchmarks indicate that the proposed method achieves very good performances on a variety of scenarios. An assessment on long video streams confirms our method incorporates periodical changes, as the ones caused by variations in natural illumination. The model, fully data driven, is suitable as a main component of a change detection system.

We present an algorithm for dictionary learning that is based on the alternating proximal algorithm studied by Attouch, Bolte, Redont, and Soubeyran (2010), coupled with a reliable and efficient dual algorithm for computation of the related proximity operators. This algorithm is suitable for a general dictionary learning model composed of a Bregman-type data fit term that accounts for the goodness of the representation and several convex penalization terms on the coefficients and atoms, explaining the prior knowledge at hand. As Attouch et al. recently proved, an alternating proximal scheme ensures better convergence properties than the simpler alternating minimization. We take care of the issue of inexactness in the computation of the involved proximity operators, giving a sound stopping criterion for the dual inner algorithm, which keeps under control the related errors, unavoidable for such a complex penalty terms, providing ultimately an overall effective procedure. Thanks to the generality of the proposed framework, we give an application in the context of genome-wide data understanding, revising the model proposed by Nowak, Hastie, Pollack, and Tibshirani (2011). The aim is to extract latent features (atoms) and perform segmentation on array-based comparative genomic hybridization (aCGH) data. We improve several important aspects that increase the quality and interpretability of the results. We show the effectiveness of the proposed model with two experiments on synthetic data, which highlight the enhancements over the original model.

In this work we build a model of the background based on dictionary learning. The image is divided into patches of equal size and a background model is obtained as a sparse linear combination of patch prototypes learnt from the image stream and updated when necessary to take into account stable variations. By enforcing sparsity, the obtained reconstruction can be computed and maintained effectively. The proposed method is stable with respect to illumination changes, correctly incorporates stable background changes in the model, and cancels out moving objects. Experiments on benchmark data indicate that the proposed method reaches very good pixel-wise performances even if relatively large patches are used.

We deal with the problem of variable selection when variables must be selected group-wise, with possibly overlapping groups defined a priori. In particular we propose a new optimization procedure for solving the regularized algorithm presented in Jacob et al. 09, where the group lasso penalty is generalized to overlapping groups of variables. While in Jacob et al. 09 the proposed implementation requires explicit replication of the variables belonging to more than one group, our iterative procedure is based on a combination of proximal methods in the primal space and constrained Newton method in a reduced dual space, corresponding to the active groups. This procedure provides a scalable alternative with no need for data duplication, and allows to deal with high dimensional problems without pre-processing to reduce the dimensionality of the data. The computational advantages of our scheme with respect to state-of-the-art algorithms using data duplication are shown empirically with numerical simulations.

In this paper, we propose a new trainable system for selecting face features from over-complete dictionaries of image measurements. The starting point is an iterative thresholding algorithm which provides sparse solutions to linear systems of equations. Although the proposed methodology is quite general and could be applied to various image classification tasks, we focus here on the case study of face and eyes detection. For our initial representation, we adopt rectangular features in order to allow straightforward comparisons with existing techniques. For computational efficiency and memory saving requirements, instead of implementing the full optimization scheme on tenths of thousands of features, we propose a three-stage architecture which consists of finding first intermediate solutions to smaller size optimization problems, then merging the obtained results, and next applying further selection procedures. The devised system requires the solution of a number of independent problems, and, hence, the necessary computations could be implemented in parallel. Experimental results obtained on both benchmark and newly acquired face and eyes images indicate that our method is a serious competitor to other feature selection schemes recently popularized in computer vision for dealing with problems of real-time object detection. A major advantage of the proposed system is that it performs well even with relatively small training sets.

We discuss how a large class of regularization methods, collectively known as spectral regularization and originally designed for solving ill-posed inverse problems, gives rise to regularized learning algorithms. All of these algorithms are consistent kernel methods that can be easily implemented. The intuition behind their derivation is that the same principle allowing for the numerical stabilization of a matrix inversion problem is crucial to avoid overfitting. The various methods have a common derivation but different computational and theoretical properties. We describe examples of such algorithms, analyze their classification performance on several data sets and discuss their applicability to real-world problems.

We describe a trainable system for face detection and tracking. The structure of the system is based on multiple cues that discard non face areas as soon as possible: we combine motion, skin, and face detection. The latter is the core of our system and consists of a hierarchy of small SVM classifiers built on the output of an automatic feature selection procedure. Our feature selection is entirely data-driven and allows us to obtain powerful descriptions from a relatively small set of data. Finally, a Kalman tracking on the face region optimizes detection results over time. We present an experimental analysis of the face detection module and results obtained with the whole system on the specific task of counting people entering the scene.

In this paper we investigate the regularization property of Kernel Principal Component Analysis (KPCA), by studying its application as a preprocessing step to supervised learning problems. We show that performing KPCA and then ordinary least squares on the projected data, a procedure known as kernel principal component regression (KPCR), is equivalent to spectral cut-off regularization, the regularization parameter being exactly the number of principal components to keep. Using probabilistic estimates for integral operators we can prove error estimates for KPCR and propose a parameter choice procedure allowing to prove consistency of the algorithm.

In this paper we discuss a general framework for feature selection based on nonparametric statistics. The three stage approach we propose is based on the assumption that the available data set is representative of a certain concept and aims at learning from the data the selection of a subset of descriptive features out of a large pool of measurements. The first stage requires the computation of a large number of image features. Simple significance tests and the maximum likelihood principle are at the basis of the second stage in which a saliency measure is used to reject the features which do not appear to be descriptive of the given data set. The third and final stage, by using the Spearman independence rank test, selects a maximal number of pairwise independent features. We report experiments on a face dataset (the MIT-CBCL database) which confirm the quality and the potential of the approach.

Kernel Methods are algorithms that, by replacing the inner product with an appropriate positive definite function, implicitly perform a nonlinear mapping of the input data into a high-dimensional feature space. In this paper, we present a kernel method for clustering inspired by the classical K-Means algorithm in which each cluster is iteratively refined using a one-class Support Vector Machine. Our method, which can be easily implemented, compares favorably with respect to popular clustering algorithms, like K-Means, Neural Gas, and Self-Organizing Maps, on a synthetic data set and three UCI real data benchmarks (IRIS data, Wisconsin breast cancer database, Spam database).

In the statistical learning framework, the use of appropriate kernels may be the key for substantial improvement in solving a given problem. In essence, a kernel is a similarity measure between input points satisfying some mathematical requirements and possibly capturing the domain knowledge. In this paper, we focus on kernels for images: we represent the image information content with binary strings and discuss various bitwise manipulations obtained using logical operators and convolution with nonbinary stencils. In the theoretical contribution of our work, we show that histogram intersection is a Mercer's kernel and we determine the modifications under which a similarity measure based on the notion of Hausdorff distance is also a Mercer's kernel. In both cases, we determine explicitly the mapping from input to feature space. The presented experimental results support the relevance of our analysis for developing effective trainable systems.

In this letter, we investigate the impact of choosing different loss functions from the viewpoint of statistical learning theory. We introduce a convexity assumption, which is met by all loss functions commonly used in the literature, and study how the bound on the estimation error changes with the loss. We also derive a general result on the minimizer of the expected risk for a convex loss function in the case of classification. The main outcome of our analysis is that for classification, the hinge loss appears to be the loss of choice. Other things being equal, the hinge loss leads to a convergence rate practically indistinguishable from the logistic loss rate and much better than the square loss rate. Furthermore, if the hypothesis space is sufficiently rich, the bounds obtained for the hinge loss are not loosened by the thresholding stage.

In this paper we address the problem of classifying images, by exploiting global features that describe color and illumination properties, and by using the statistical learning paradigm. The contribution of this paper is twofold. First, we show that histogram intersection has the required mathematical properties to be used as a kernel function for support vector machines (SVMs). Second, we give two examples of how a SVM, equipped with such a kernel, can achieve very promising results on image classification based on color information.

The problem of learning is arguably at the very core of the problem of intelligence, both biological and artificial. In this paper we review our approach to the problem of visual perception based on supervised learning. After a brief presentation of the theoretical background, we focus on some of the engineering applications of statistical learning to computer vision and discuss the main open problems and directions of our future research.

We investigate the automatic estimation of fish weight from sets of morphometric measurements. Our solution combines a vision system with a robust regression method, the Support Vector Machine (SVM). Measurements are taken automatically from two binarised views of each fish in a training sample, then fed to a quadratic SVM along with approximate weight estimates. The SVM learns the law linking weight to shape directly (without computing volume) and compensates for several inaccuracies in the training measurements. We suggest a methodology identifying optimal shape measurements for the task, and report results obtained with a sample of 99 trouts between 300 and 600g, showing good accuracy and reliability, and better performance with respect to length-weight relations adopted commonly in fisheries science. 1 Introduction This work explores a new way of estimating fish weight from shape using computer vision. The relation between weight and shape is important both for fish biology [4, 5...

This paper presents a novel method for determining the location of the instantaneous epipole in a sequence of images acquired by an uncalibrated camera and containing a single, rigid motion (e.g., the camera moves in a static environment). The method uses the full perspective camera model and requires the estimation of the optical flow at a minimum of six image locations. The key observation is that the optical flow equations can be written in terms of the epipole in a strikingly simple form if the translational and rotational flow components are not separated as done usually. The epipole location can then be obtained as the minimum of a least-square residual function associated to the computed optical flow. We report and discuss initial experiments on both synthetic and real data and illustrate possible developments of this method towards the use of uncalibrated optical flow for 3-D motion and structure reconstruction.

Support vector machines (SVMs) perform pattern recognition between two point classes by finding a decision surface determined by certain points of the training set, termed support vectors (SV). This surface, which in some feature space of possibly infinite dimension can be regarded as a hyperplane, is obtained from the solution of a problem of quadratic programming that depends on a regularization parameter. In this article, we study some mathematical properties of support vectors and show that the decision surface can be written as the sum of two orthogonal terms, the first depending on only the margin vectors (which are SVs lying on the margin), the second proportional to the regularization parameter. For almost all values of the parameter, this enables us to predict how the decision surface varies for small parameter changes. In the special but important case of feature space of finite dimension m, we also show that m + 1 SVs are usually sufficient to determine the decision surface fully. For relatively small m, this latter result leads to a consistent reduction of the SV number.

Support vector machines (SVMs) have been recently proposed as a new technique for pattern recognition. Intuitively, given a set of points which belong to either of two classes, a linear SVM finds the hyperplane leaving the largest possible fraction of points of the same class on the same side, while maximizing the distance of either class from the hyperplane. The hyperplane is determined by a subset of the points of the two classes, named support vectors, and has a number of interesting theoretical properties. In this paper, we use linear SVMs for 3D object recognition. We illustrate the potential of SVMs on a database of 7200 images of 100 different objects. The proposed system does not require feature extraction and performs recognition on images regarded as points of a space of high dimension without estimating pose. The excellent recognition rates achieved in all the performed experiments indicate that SVMs are well-suited for aspect-based recognition.

An aspect-based system for the recognition of 3D objects from single view is presented. The system is based on the computation of size functions and consists of two stages: 1) models of the various aspects of the objects in a set are acquired from the corresponding edge maps, each model is represented by a feature vector and a training set is formed; and 2) a feature vector representing the shape of an object from a single previously unseen image is constructed and classified according to a k-nearest neighbour technique. The system was tested on a set of thirteen toy cars arbitrarily positioned on a turntable and viewed from a fixed, uncalibrated camera, and compared against methods based on moments (MB) and on Hausdorff distance (HDB). Since the system outperforms MB methods in terms of percentages of success and the HDB method in terms of efficiency, it is concluded that size functions can be very useful for aspect-based recognition.

This paper discusses the use of certain integer valued functions of two real variables, named size functions, for shape representation and recognition. The recognition of the signing alphabet is described as a study case. A number of size functions are computed from the edge map of the viewed sign and a feature vector based on the obtained size functions is formed. A training set of feature vectors built from real images and the ^-nearest-neighbor rule are employed for the classification of unpreviously seen signs. The proposed system performs recognition at about 2Hz with feature vectors of small dimension. The reported experiments indicate that size functions can be effectively used for the recognition of nonrigid shapes. 1

Size functions are integer-valued functions which represent both qualitative and quantitative properties of visual shape. In this paper the use of size functions for the understanding of the alphabet of sign language is described. First, a family of size functions able to capture important aspects of shape from the apparent outline of the various signs are presented and motivated. Each sign is represented by means of a feature vector computed from the proposed family of size functions. Then, a training set of feature vectors is built from real images. Finally, the k-nearest-neighbor rule is employed for the classification of feature vectors computed from previously unseen signs. The reported experiments indicate that size functions can be extremely effective for the recognition of signs even in the presence of shape changes due to difference in hands, pose, style of signing, and viewpoint.

A method is presented for the recovery of optical flow. The key idea is that the local spatial structure of optical flow, with the exception of surface boundaries, is usually rather coherent and can thus be appropriately approximated by a linear vector field. According to the proposed method, the optical flow components and their first order spatial derivatives are computed at the central points of rather large and overlapping patches which cover the image plane as the solution to a highly overconstrained system of linear algebraic equations. The equations, which are solved through the use of standard least mean square techniques, are derived from the assumptions that the changing image brightness is stationary everywhere over time and that optical flow is, locally, a linear vector field. The method has been tested on many sequences of synthetic and real images and the obtained optical flow has been used to estimate three-dimensional motion parameters with very good results.>

It is shown that the motion field the 2-D vector field which is the perspective projection on the image plane of the 3-D velocity field of a moving scene, and the optical flow, defined as the estimate of the motion field which can be derived from the first-order variation of the image brightness pattern, are in general different, unless special conditions are satisfied. Therefore, dense optical flow is often ill-suited for computing structure from motion and for reconstructing the 3-D velocity field by algorithms which require a locally accurate estimate of the motion field. A different use of the optical flow is suggested. It is shown that the (smoothed) optical flow and the motion field can be interpreted as vector fields tangent to flows of planar dynamical systems. Stable qualitative properties of the motion field, which give useful informations about the 3-D velocity field and the 3-D structure of the scene, usually can be obtained from the optical flow. The idea is supported by results from the theory of structural stability of dynamical systems.>