Search

This paper introduces the Neural Transcoding Vision Transformer (\modelname), a generative model designed to estimate high-resolution functional Magnetic Resonance Imaging (fMRI) samples from simultaneous Electroencephalography (EEG) data. A key feature of \modelname is its Domain Matching (DM) sub-module which effectively aligns the latent EEG representations with those of fMRI volumes, enhancing the model's accuracy and reliability. Unlike previous methods that tend to struggle with fidelity and reproducibility of images, \modelname addresses these challenges by ensuring methodological integrity and higher-quality reconstructions which we showcase through extensive evaluation on two benchmark datasets; \modelname outperforms the current state-of-the-art by a significant margin in both cases, e.g. achieving a $10\times$ reduction in RMSE and a $3.14\times$ increase in SSIM on the Oddball dataset. An ablation study also provides insights into the contribution of each component to the model's overall effectiveness. This development is critical in offering a new approach to lessen the time and financial constraints typically linked with high-resolution brain imaging, thereby aiding in the swift and precise diagnosis of neurological disorders. Although it is not a replacement for actual fMRI but rather a step towards making such imaging more accessible, we believe that it represents a pivotal advancement in clinical practice and neuroscience research. Code is available at \url{https://github.com/rom42pla/ntvit}., Comment: ECCV24 Workshop on Synthetic Data for Computer Vision

Action anticipation is the task of forecasting future activity from a partially observed sequence of events. However, this task is exposed to intrinsic future uncertainty and the difficulty of reasoning upon interconnected actions. Unlike previous works that focus on extrapolating better visual and temporal information, we concentrate on learning action representations that are aware of their semantic interconnectivity based on prototypical action patterns and contextual co-occurrences. To this end, we propose the novel Semantically Guided Representation Learning (S-GEAR) framework. S-GEAR learns visual action prototypes and leverages language models to structure their relationship, inducing semanticity. To gather insights on S-GEAR's effectiveness, we test it on four action anticipation benchmarks, obtaining improved results compared to previous works: +3.5, +2.7, and +3.5 absolute points on Top-1 Accuracy on Epic-Kitchen 55, EGTEA Gaze+ and 50 Salads, respectively, and +0.8 on Top-5 Recall on Epic-Kitchens 100. We further observe that S-GEAR effectively transfers the geometric associations between actions from language to visual prototypes. Finally, S-GEAR opens new research frontiers in anticipation tasks by demonstrating the intricate impact of action semantic interconnectivity., Comment: Accepted as a full paper at ECCV'24 with Paper ID #4140

Deepfake detection aims to contrast the spread of deep-generated media that undermines trust in online content. While existing methods focus on large and complex models, the need for real-time detection demands greater efficiency. With this in mind, unlike previous work, we introduce a novel deepfake detection approach on images using Binary Neural Networks (BNNs) for fast inference with minimal accuracy loss. Moreover, our method incorporates Fast Fourier Transform (FFT) and Local Binary Pattern (LBP) as additional channel features to uncover manipulation traces in frequency and texture domains. Evaluations on COCOFake, DFFD, and CIFAKE datasets demonstrate our method's state-of-the-art performance in most scenarios with a significant efficiency gain of up to a $20\times$ reduction in FLOPs during inference. Finally, by exploring BNNs in deepfake detection to balance accuracy and efficiency, this work paves the way for future research on efficient deepfake detection., Comment: Accepted at CVPR24 DFAD Workshop

In this paper, we i) analyze and classify real-world failures of Kubernetes (the most popular container orchestration system), ii) develop a framework to perform a fault/error injection campaign targeting the data store preserving the cluster state, and iii) compare results of our fault/error injection experiments with real-world failures, showing that our fault/error injections can recreate many real-world failure patterns. The paper aims to address the lack of studies on systematic analyses of Kubernetes failures to date. Our results show that even a single fault/error (e.g., a bit-flip) in the data stored can propagate, causing cluster-wide failures (3% of injections), service networking issues (4%), and service under/overprovisioning (24%). Errors in the fields tracking dependencies between object caused 51% of such cluster-wide failures. We argue that controlled fault/error injection-based testing should be employed to proactively assess Kubernetes' resiliency and guide the design of failure mitigation strategies.

With the increasing use of multicore platforms to realize mixed-criticality systems, understanding the underlying shared resources, such as the memory hierarchy shared among cores, and achieving isolation between co-executing tasks running on the same platform with different criticality levels becomes relevant. In addition to safety considerations, a malicious entity can exploit shared resources to create timing attacks on critical applications. In this paper, we focus on understanding the shared DRAM dual in-line memory module and created a timing attack, that we named the "bank & row conflict bomb", to target a victim task in a multicore platform. We also created a "navigate" algorithm to understand how victim requests are managed by the Memory Controller and provide valuable inputs for designing the bank & row conflict bomb. We performed experimental tests on a 2nd Gen Intel Xeon Processor with an 8GB DDR4-2666 DRAM module to show that such an attack can produce a significant increase in the execution time of the victim task by about 150%, motivating the need for proper countermeasures to help ensure the safety and security of critical applications., Comment: To appear in the proceedings of the 27th IEEE international Symposium on Real-time Distributed Computing (ISORC)

The adoption of cloud computing technologies in the industry is paving the way to new manufacturing paradigms. In this paper we propose a model to optimize the orchestration of workloads with differentiated criticality levels on a cloud-enabled factory floor. Preliminary results show that it is possible to optimize the guarantees to deployed jobs without penalizing the number of schedulable jobs. We indicate future research paths to quantitatively evaluate job isolation., Comment: S. Bernardi, T. Zoppi (Editors), Fast Abstracts and Student Forum Proceedings - EDCC 2024 - 19th European Dependable Computing Conference, Leuven, Belgium, 8-11 April 2024

We study Cauchy problems of fractional differential equations in both space and time variables by expressing the solution in terms of ``stochastic composition" of the solutions to two simpler problems. These Cauchy sub-problems respectively concern the space and the time differential operator involved in the main equation. We provide some probabilistic and pseudo-probabilistic applications, where the solution can be interpreted as the pseudo-transition density of a time-changed pseudo-process. To extend our results to higher order time-fractional problems, we introduce pseudo-subordinators as well as its pseudo-inverse. Finally, we present our results in the case of more general differential operators and we interpret the results by means of a linear combination of pseudo-subordinators and its inverse., Comment: 18

Vision Transformer (ViT) self-attention mechanism is characterized by feature collapse in deeper layers, resulting in the vanishing of low-level visual features. However, such features can be helpful to accurately represent and identify elements within an image and increase the accuracy and robustness of vision-based recognition systems. Following this rationale, we propose a novel residual attention learning method for improving ViT-based architectures, increasing their visual feature diversity and model robustness. In this way, the proposed network can capture and preserve significant low-level features, providing more details about the elements within the scene being analyzed. The effectiveness and robustness of the presented method are evaluated on five image classification benchmarks, including ImageNet1k, CIFAR10, CIFAR100, Oxford Flowers-102, and Oxford-IIIT Pet, achieving improved performances. Additionally, experiments on the COCO2017 dataset show that the devised approach discovers and incorporates semantic and spatial relationships for object detection and instance segmentation when implemented into spatial-aware transformer models., Comment: 17 pages, single column, 9 figures

Inhomogeneous cloud formation and wavelength-dependent phenomena are expected to shape hot Jupiter atmospheres. We present a General Circulation Model (GCM) with multiwavelength "picket fence" radiative transfer and radiatively active, temperature dependent clouds, and compare the results to a double gray routine. The double gray method inherently fails to model polychromatic effects in hot Jupiter atmospheres, while picket fence captures these non-gray aspects and performs well compared to fully wavelength-dependent methods. We compare both methods with radiatively active clouds and cloud-free models, assessing the limitations of the double gray method. Although there are broad similarities, the picket fence models have larger day-night side temperature differences, non-isothermal upper atmospheres, and multiwavelength effects in the presence of radiatively active clouds. We model the well-known hot Jupiters HD 189733 b and HD 209458 b. For the hotter HD 209458 b, the picket fence method prevents clouds from thermostating dayside temperatures, resulting in hotter upper atmospheres and the dissipation of dayside clouds. Differences in the temperature structures are then associated with nuanced differences in the circulation patterns and clouds. Models of the cooler HD 189733 b have global cloud coverage, regardless of radiative transfer scheme, whereas there are larger differences in the models of HD 209458 b, particularly in the extent of the partial cloud coverage on its dayside. This results in minor changes to the thermal and reflected light phase curves of HD 189733 b, but more significant differences for the picket fence and double gray versions of HD 209458 b., Comment: Accepted in ApJ, 31 pages

Hemoglycin, a space polymer of glycine and iron, has been identified in the carbonaceous chondritic meteorites Allende, Acfer 086, Kaba, Sutters Mill and Orgueil. Its core form has a mass of 1494Da and is basically an antiparallel pair of polyglycine strands linked at each end by an iron atom. The polymer forms two- and three- dimensional lattices with an inter-vertex distance of 4.9nm. Here the extraction technique for meteorites is applied to a 2.1Gya fossil stromatolite to reveal the presence of hemoglycin by mass spectrometry. Intact ooids from a recent (3,000Ya) stromatolite exhibited the same visible hemoglycin fluorescence in response to x-rays as an intact crystal from the Orgueil meteorite. X-ray analysis confirmed the existence in ooids of an internal 3-dimensional lattice of 4.9nm inter-vertex spacing, matching the spacing of lattices in meteoritic crystals. FTIR measurements of acid-treated ooid and a Sutters Mill meteoritic crystal both show the presence, via the splitting of the Amide I band, of an extended anti-parallel beta sheet structure. It seems probable that the copious in-fall of carbonaceous meteoritic material, from Archaean times onward, has left traces of hemoglycin in sedimentary carbonates and potentially has influenced ooid formation., Comment: 45 pages 16 Figure 8 Tables

In this paper we give an explicit solution of Dzherbashyan-Caputo-fractional Cauchy problems related to equations with derivatives of order $\nu k$, for $k$ non-negative integer and $\nu>0$. The solution is obtained by connecting the differential equation with the roots of the characteristic polynomial and it is expressed in terms of Mittag-Leffler-type functions. Under the some stricter hypothesis the solution can be expressed as a linear combination of Mittag-Leffler functions with common fractional order $\nu$. We establish a probabilistic relationship between the solutions of differential problems with order $\nu/m$ and $\nu$, for natural $m$. Finally, we use the described method to solve fractional differential equations arising in the fractionalization of partial differential equations related to the probability law of planar random motions with finite velocities., Comment: 21 pages

We present a detailed analysis of random motions moving in higher spaces with a natural number of velocities. In the case of the so-called minimal random dynamics, under some wide assumptions, we show the joint distribution of the position of the motion (both for the inner part and the border of the support) and the number of displacements performed with each velocity. Explicit results for cyclic and complete motions are derived. We establish useful relationships between motions moving in different spaces and we derive the form of the distribution of the movements in arbitrary dimension. Finally, we investigate further properties for stochastic motions governed by non-homogeneous Poisson processes., Comment: 29 pages

Nowadays, industries are looking into virtualization as an effective means to build safe applications, thanks to the isolation it can provide among virtual machines (VMs) running on the same hardware. In this context, a fundamental issue is understanding to what extent the isolation is guaranteed, despite possible (or induced) problems in the virtualization mechanisms. Uncovering such isolation issues is still an open challenge, especially for hardware-assisted virtualization, since the search space should include all the possible VM states (and the linked hypervisor state), which is prohibitive. In this paper, we propose IRIS, a framework to record (learn) sequences of inputs (i.e., VM seeds) from the real guest execution (e.g., OS boot), replay them as-is to reach valid and complex VM states, and finally use them as valid seed to be mutated for enabling fuzzing solutions for hardware-assisted hypervisors. We demonstrate the accuracy and efficiency of IRIS in automatically reproducing valid VM behaviors, with no need to execute guest workloads. We also provide a proof-of-concept fuzzer, based on the proposed architecture, showing its potential on the Xen hypervisor., Comment: 13 pages, Accepted for publication at The 53rd Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN)

Though the global atmospheres of hot Jupiters have been extensively studied using phase curve observations, the level of time variability in these data is not well constrained. To investigate possible time variability in a planetary phase curve, we observed two full-orbit phase curves of the hot Jupiter WASP-43b at 4.5 microns using the Spitzer Space Telescope, and reanalyzed a previous 4.5 micron phase curve from Stevenson et al. (2017). We find no significant time variability between these three phase curves, which span timescales of weeks to years. The three observations are best fit by a single phase curve with an eclipse depth of 3907 +- 85 ppm, a dayside-integrated brightness temperature of 1479 +- 13 K, a nightside-integrated brightness temperature of 755 +- 46 K, and an eastward-shifted peak of 10.4 +- 1.8 degrees. To model our observations, we performed 3D general circulation model simulations of WASP-43b with simple cloud models of various vertical extents. In comparing these simulations to our observations, we find that WASP-43b likely has a cloudy nightside that transitions to a relatively cloud-free dayside. We estimate that any change in WASP-43bs vertical cloud thickness of more than three pressure scale heights is inconsistent with our observed upper limit on variation. These observations, therefore, indicate that WASP-43bs clouds are stable in their vertical and spatial extent over timescales up to several years. These results strongly suggest that atmospheric properties derived from previous, single Spitzer phase curve observations of hot Jupiters likely show us the equilibrium properties of these atmospheres., Comment: 24 pages, 9 figures, Published in the Astronomical Journal (AJ)

The goal of this paper is to introduce pooling strategies for simplicial convolutional neural networks. Inspired by graph pooling methods, we introduce a general formulation for a simplicial pooling layer that performs: i) local aggregation of simplicial signals; ii) principled selection of sampling sets; iii) downsampling and simplicial topology adaptation. The general layer is then customized to design four different pooling strategies (i.e., max, top-k, self-attention, and separated top-k) grounded in the theory of topological signal processing. Also, we leverage the proposed layers in a hierarchical architecture that reduce complexity while representing data at different resolutions. Numerical results on real data benchmarks (i.e., flow and graph classification) illustrate the advantage of the proposed methods with respect to the state of the art.

Orchestration systems are becoming a key component to automatically manage distributed computing resources in many fields with criticality requirements like Industry 4.0 (I4.0). However, they are mainly linked to OS-level virtualization, which is known to suffer from reduced isolation. In this paper, we propose RunPHI with the aim of integrating partitioning hypervisors, as a solution for assuring strong isolation, with OS-level orchestration systems. The purpose is to enable container orchestration in mixed-criticality systems with isolation requirements through partitioned containers., Comment: 2 pages, accepted for publication in Proc. ISSREW, 2022

Industries are recently considering the adoption of cloud computing for hosting safety critical applications. However, the use of multicore processors usually adopted in the cloud introduces temporal anomalies due to contention for shared resources, such as the memory subsystem. In this paper we explore the potential of Intel's Memory Bandwidth Allocation (MBA) technology, available on Xeon Scalable processors. By adopting a systematic measurement approach on real hardware, we assess the indirect memory bandwidth limitation achievable by applying MBA delays, showing that only given delay values (namely 70, 80 and 90) are effective in our setting. We also test the derived bandwidth assured to a hypothetical critical core when interfering cores (e.g., generating a concurrent memory access workload) are present on the same machine. Our results can support designers by providing understanding of impact of the shared memory to enable predictable progress of safety critical applications in cloud environments., Comment: 8 pages, to appear in proceedings of "The 25th International Symposium On Real-Time Distributed Computing ISORC"

The purpose was two-fold: To explore the capability of the Diamond Synchrotron infra-red so as to include near-field photothermal microspectroscopy (PTMS); and Toward a quantitative nanohistology - investigation of scleroderma using synchrotron radiation (mu- FTIR). With recent advances in AFM, the integration of an IR temperature-based system on an IR beamline is still a promising and unique approach. These preliminary tests of the PTMS stand-alone system with the Diamond Multimode InfraRed Imaging and Microspectroscopy beamline (MIRIAM) were successful in one beamtime. To our knowledge, this is still among very few studies worldwide that managed to interface an AFM-resistive probe to measure an infrared spectrum using synchrotron radiation (SR) as a source. When the alignment was successful, we managed to obtain good spectra on samples that were strong infrared absorbers. However, the alignment could not be guaranteed all the time, and the signal-to-noise ratio became too low to record any meaningful spectra on samples deemed of scientific interest for this proposal. We clarified the possible improvements to PTMS instrumentation and data acquisition needed to ensure that we can collect sufficient thermal signals from very thin samples., Comment: 10 pages.12 figures, two additional references with relevant comments added to text

Time predictable edge cloud is seen as the answer for many arising needs in Industry 4.0 environments, since it is able to provide flexible, modular, and reconfigurable services with low latency and reduced costs. Orchestration systems are becoming the core component of clouds since they take decisions on the placement and lifecycle of software components. Current solutions start introducing real-time containers support for time predictability; however, these approaches lack of determinism as well as support for workloads requiring multiple levels of assurance/criticality. In this paper, we present k4.0s, an orchestration model for real-time and mixed-criticality environments, which includes timeliness, criticality and network requirements. The model leverages new abstractions for both node and jobs, e.g., node assurance, and requires novel monitoring strategies. We sketch an implementation of the proposal based on Kubernetes, and present an experimentation motivating the need for node assurance levels and adequate monitoring., Comment: This paper is an extended version of the paper published at "2022 IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology Congress"

Nowadays, a feature-rich automotive vehicle offers several technologies to assist the driver during his trip and guarantee an amusing infotainment system to the other passengers, too. Consolidating worlds at different criticalities is a welcomed challenge for car manufacturers that have recently tried to leverage virtualization technologies due to reduced maintenance, deployment, and shipping costs. For this reason, more and more mixed-criticality systems are emerging, trying to assure compliance with the ISO 26262 Road Vehicle Safety standard. In this short paper, we provide a preliminary investigation of the certification capabilities for Jailhouse, a popular open-source partitioning hypervisor. To this aim, we propose a testing methodology and showcase the results, pointing out when the software gets to a faulting state, deviating from its expected behavior. The ultimate goal is to picture the right direction for the hypervisor towards a potential certification process., Comment: 4 pages, to be published in "8th international workshop on safety and security of intelligentt vehicles" [SSIV 2022]

This paper is devoted to the detailed analysis of three-dimensional motions in $\mathbb{R}^3$ with orthogonal directions switching at Poisson times and moving with constant speed $c>0$. The study of the random position at an arbitrary time $t>0$ on the surface of the support, forming an octahedron $S_{ct}$, is completely carried out on the edges $E_{ct}$ and faces $F_{ct}$. In particular, the motion on the faces $F_{ct}$ is analysed by means of a transformation which reduces it to a three-directions planar random motion. This permits us to obtain an integral representation on $F_{ct}$ in terms of integral of products of first order Bessel functions. The investigation of the distribution of the position $p=p(t,x,y,z)$ inside $S_{ct}$ implied the derivation of a sixth-order partial differential equation governing $p$ (expressed in terms of the products of three D'Alembert operators). A number of results, also in explicit form, concern the time spent on each direction and the position reached by each coordinates as the motion devolpes. The analysis is carried out when the incoming direction is orthogonal to the ongoing one and also when all directions can be uniformely choosen at each Poisson event. If the switches are governed by homogeneus Poisson process many explicit results are obtained.

Taking medicines is a fundamental aspect to cure illnesses. However, studies have shown that it can be hard for patients to remember the correct posology. More aggravating, a wrong dosage generally causes the disease to worsen. Although, all relevant instructions for a medicine are summarized in the corresponding patient information leaflet, the latter is generally difficult to navigate and understand. To address this problem and help patients with their medication, in this paper we introduce an augmented reality mobile application that can present to the user important details on the framed medicine. In particular, the app implements an inference engine based on a deep neural network, i.e., a densenet, fine-tuned to recognize a medicinal from its package. Subsequently, relevant information, such as posology or a simplified leaflet, is overlaid on the camera feed to help a patient when taking a medicine. Extensive experiments to select the best hyperparameters were performed on a dataset specifically collected to address this task; ultimately obtaining up to 91.30\% accuracy as well as real-time capabilities., Comment: 12 pages, 7 figures

Nowadays, machine and deep learning techniques are widely used in different areas, ranging from economics to biology. In general, these techniques can be used in two ways: trying to adapt well-known models and architectures to the available data, or designing custom architectures. In both cases, to speed up the research process, it is useful to know which type of models work best for a specific problem and/or data type. By focusing on EEG signal analysis, and for the first time in literature, in this paper a benchmark of machine and deep learning for EEG signal classification is proposed. For our experiments we used the four most widespread models, i.e., multilayer perceptron, convolutional neural network, long short-term memory, and gated recurrent unit, highlighting which one can be a good starting point for developing EEG classification models., Comment: conference, 11 pages, 5 figures

The increasing availability of wireless access points (APs) is leading towards human sensing applications based on Wi-Fi signals as support or alternative tools to the widespread visual sensors, where the signals enable to address well-known vision-related problems such as illumination changes or occlusions. Indeed, using image synthesis techniques to translate radio frequencies to the visible spectrum can become essential to obtain otherwise unavailable visual data. This domain-to-domain translation is feasible because both objects and people affect electromagnetic waves, causing radio and optical frequencies variations. In literature, models capable of inferring radio-to-visual features mappings have gained momentum in the last few years since frequency changes can be observed in the radio domain through the channel state information (CSI) of Wi-Fi APs, enabling signal-based feature extraction, e.g., amplitude. On this account, this paper presents a novel two-branch generative neural network that effectively maps radio data into visual features, following a teacher-student design that exploits a cross-modality supervision strategy. The latter conditions signal-based features in the visual domain to completely replace visual data. Once trained, the proposed method synthesizes human silhouette and skeleton videos using exclusively Wi-Fi signals. The approach is evaluated on publicly available data, where it obtains remarkable results for both silhouette and skeleton videos generation, demonstrating the effectiveness of the proposed cross-modality supervision strategy.

We consider the telegraph process with two velocities, $a_1>a_2\in\mathbb{R}$, and two rates of reversal, $\lambda_1,\lambda_2>0$. We study some of its features with respect to the conditional probability measure where both the initial speed and the number of changes of direction are known. We exhibit a new proof by induction of the (conditional) probability law and a detailed study of the distribution of the motion at time $t>0$ conditioned on its position at a previous time $00$, its maximum and its minimum up to that moment., Comment: 15 pages, 2 figures

The next-generation Industrial Internet of Things (IIoT) inherently requires smart devices featuring rich connectivity, local intelligence, and autonomous behavior. Emerging Multiprocessor System-on-Chip (MPSoC) platforms along with comprehensive support for virtualization will represent two key building blocks for smart devices in future IIoT edge infrastructures. We review representative existing solutions, highlighting the aspects that are most relevant for integration in IIoT solutions. From the analysis, we derive a reference architecture for a general virtualization-ready edge IIoT node. We then analyze the implications and benefits for a concrete use case scenario and identify the crucial research challenges to be faced to bridge the gap towards full support for virtualization-ready IIoT nodes, Comment: 10 pages, 3 figures, 1 table, accepted for publication in IEEE Computer

Virtualization is gaining attraction in the industry as it promises a flexible way to integrate, manage, and re-use heterogeneous software components with mixed-criticality levels, on a shared hardware platform, while obtaining isolation guarantees. This work surveys the state-of-the-practice of real-time virtualization technologies by discussing common issues in the industry. In particular, we analyze how different virtualization approaches and solutions can impact isolation guarantees and testing/certification activities, and how they deal with dependability challenges. The aim is to highlight current industry trends and support industrial practitioners to choose the most suitable solution according to their application domains., Comment: Accepted for publication in Elsevier Future Generation Computer Systems

Over the last year, the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and its variants have highlighted the importance of screening tools with high diagnostic accuracy for new illnesses such as COVID-19. To that regard, deep learning approaches have proven as effective solutions for pneumonia classification, especially when considering chest-x-rays images. However, this lung infection can also be caused by other viral, bacterial or fungi pathogens. Consequently, efforts are being poured toward distinguishing the infection source to help clinicians to diagnose the correct disease origin. Following this tendency, this study further explores the effectiveness of established neural network architectures on the pneumonia classification task through the transfer learning paradigm. To present a comprehensive comparison, 12 well-known ImageNet pre-trained models were fine-tuned and used to discriminate among chest-x-rays of healthy people, and those showing pneumonia symptoms derived from either a viral (i.e., generic or SARS-CoV-2) or bacterial source. Furthermore, since a common public collection distinguishing between such categories is currently not available, two distinct datasets of chest-x-rays images, describing the aforementioned sources, were combined and employed to evaluate the various architectures. The experiments were performed using a total of 6330 images split between train, validation and test sets. For all models, common classification metrics were computed (e.g., precision, f1-score) and most architectures obtained significant performances, reaching, among the others, up to 84.46% average f1-score when discriminating the 4 identified classes. Moreover, confusion matrices and activation maps computed via the Grad-CAM algorithm were also reported to present an informed discussion on the networks classifications.

Improving existing neural network architectures can involve several design choices such as manipulating the loss functions, employing a diverse learning strategy, exploiting gradient evolution at training time, optimizing the network hyper-parameters, or increasing the architecture depth. The latter approach is a straightforward solution, since it directly enhances the representation capabilities of a network; however, the increased depth generally incurs in the well-known vanishing gradient problem. In this paper, borrowing from different methods addressing this issue, we introduce an interlaced multi-task learning strategy, defined SIRe, to reduce the vanishing gradient in relation to the object classification task. The presented methodology directly improves a convolutional neural network (CNN) by preserving information from the input image through interlaced auto-encoders (AEs), and further refines the base network architecture by means of skip and residual connections. To validate the presented methodology, a simple CNN and various implementations of famous networks are extended via the SIRe strategy and extensively tested on five collections, i.e., MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and Caltech-256; where the SIRe-extended architectures achieve significantly increased performances across all models and datasets, thus confirming the presented approach effectiveness.

Estimating the 3D pose of a hand from a 2D image is a well-studied problem and a requirement for several real-life applications such as virtual reality, augmented reality, and hand gesture recognition. Currently, reasonable estimations can be computed from single RGB images, especially when a multi-task learning approach is used to force the system to consider the shape of the hand when its pose is determined. However, depending on the method used to represent the hand, the performance can drop considerably in real-life tasks, suggesting that stable descriptions are required to achieve satisfactory results. In this paper, we present a keypoint-based end-to-end framework for 3D hand and pose estimation and successfully apply it to the task of hand gesture recognition as a study case. Specifically, after a pre-processing step in which the images are normalized, the proposed pipeline uses a multi-task semantic feature extractor generating 2D heatmaps and hand silhouettes from RGB images, a viewpoint encoder to predict the hand and camera view parameters, a stable hand estimator to produce the 3D hand pose and shape, and a loss function to guide all of the components jointly during the learning phase. Tests were performed on a 3D pose and shape estimation benchmark dataset to assess the proposed framework, which obtained state-of-the-art performance. Our system was also evaluated on two hand-gesture recognition benchmark datasets and significantly outperformed other keypoint-based approaches, indicating that it is an effective solution that is able to generate stable 3D estimates for hand pose and shape.

Collection of manuscripts accepted for presentation at the Student Forum and Fast Abstracts tracks of the 17th European Dependable Computing Conference (EDCC 2021)., Comment: Editors: Marcello Cinque and Barbara Gallina

To date, several automated strategies for identification/segmentation of Multiple Sclerosis (MS) lesions with the use of Magnetic Resonance Imaging (MRI) have been presented, but they are outperformed by human experts, from whom they act very differently. This is mainly due to: the ambiguity originated by MRI instabilities; peculiar variability of MS; non specificity of MRI regarding MS. Physicians partially manage the uncertainty generated by ambiguity relying on radiological/clinical/anatomical background and experience. To emulate human diagnosis, we present an automated framework for identification/segmentation of MS lesions from MRI based on three pivotal concepts: 1. the modelling of uncertainty; 2. the proposal of two, separately trained, CNN, one optimized for lesions and the other for lesions with respect to the environment surrounding them, respectively repeated for axial, coronal and sagittal directions; 3. the definition of an ensemble classifier to merge the information collected by different CNN. The proposed framework is trained, validated and tested on the 2016 MSSEG benchmark public data set from a single imaging modality, the FLuid-Attenuated Inversion Recovery (FLAIR). The comparison with the ground-truth and with each of 7 human raters, proves that there is no significant difference between the automated and the human raters.

We study planar random motions with finite velocities, of norm $c>0$, along orthogonal directions and changing at the instants of occurrence of a non-homogeneous Poisson process with rate function $\lambda(t),\ t\ge0$. We focus on the distribution of the current position $\bigl(X(t), Y(t)\bigr),\ t\ge0$, in the case where the motion has orthogonal deviations and where also reflection is admitted. In all the cases the process is located within the closed square $S_{ct}=\{(x,y)\in \mathbb{R}^2\,:\,|x|+|y|\le ct\}$ and we obtain the probability law inside $S_{ct}$, on the edge $\partial S_{ct}$ and on the other possible singularities, by studying the partial differential equations governing all the distributions examined. A fundamental result is that the vector process $\bigl(X(t), Y(t)\bigr)$ is probabilistically equivalent to a linear transformation of two (independent or dependent) one-dimensional symmetric telegraph processes with rate function proportional to $\lambda(t)$ and velocity $c/2$. Finally, we extend the results to a wider class of orthogonal-type evolutions.