Search

In the last decades much effort has been addressed to realize novel solid state photo-detectors with a high quantum efficiency in the UV wavelength region to be used in experiments detecting Cherenkov or fluorescence radiation even at cryogenic temperatures. Among the possible devices with these characteristics, the large-area solid detectors made of n-doped silicon substrate coated with Multi-Walled Carbon Nanotubes (MWCNTs) appear to be particularly promising since they combine the great UV radiation absorbance of MWCNTs (at about 200 nm) with their unique characteristics for electrical conductivity and mechanical resistance at low temperatures. In this work we present the cryogenic characteristics of a MWCNT-Si large-area (1 cm2) photo-detector, in which a UV photo-sensitive heterojunction is obtained growing, by Chemical Vapour Deposition (CVD), multi-walled carbon nanotubes on an n-type silicon substrate. Measurements have been made at various temperatures in the range from 5K to 300K by illuminating the photo-detector with a 378 nm UV continuous laser light source. Results demonstrate the capability of such device to be successfully employed in cryogenic experiments as well at room temperature with high stability and high photon detection efficiency in the UV region.

The tremendous evolution of the Internet has prompted the microelectronics and telecommunications industries to look at technologies that address the bandwidth, speed, and power consumption required by the Internet. One technology that is emerging as a comprehensive solution to this is integrated photonics, where complexity, reconfigurability, and intelligence are thrusts. Ideally, the best platform for such integration would be based on the same material that permitted the success of microelectronics in the last century, i.e., silicon. Silicon photonics is a technology in which photonics devices are fabricated by using standard silicon technologies.

Le moderne tecniche e metodologie di rilievo digitale 3D consentono la documentazione, conservazione e rappresentazione digitale di siti e architetture con risultati straordinari dal punti di vista geometrico e visivo. Attualmente abbiamo a nostra disposizione una vasta gamma di sensori e tecnologie per rilevare dati geometrici, processarli, derivarne informazioni 2D/3D dettagliate, accurate e metriche e infine renderle disponili attraverso basi di dati online o sistemi GIS. L'articolo presenta le attuali tecniche di rilievo per la modellazione 3D di siti culturali e architetture complesse. Esempi e problematiche sono riportati con lo scopo di valutare criticamente lo stato attuale del rilievo 3D, le potenzialità e i possibili sviluppi futuri., The actual techniques and methodologies for 3D surveying and modeling allow to achieve very satisfactory and impressive results in the field of documentation, conservation and representation of architectural sites and heritage objects. A great variety of active or passive sensors are available for data acquisition. These data are then processed in order to derive metric 3D information and share them using online databases or GIS tools. But despite the large amount of data, the processing methodologies are still relying very often on manual restitution as automatic method are still not enough reliable for the detailed and precise 3D reconstruction of architectural scenarios and drawings. The articles reports the actual 3D surveying methodologies and data processing techniques with a critical description of open problems and research issues.

Test prioritization techniques can be used to determine test case ordering and early discover faults in source code. Several of these techniques exploit a single objective function, e.g., code or requirements coverage. In this tool demo paper, we present MOTCP, a software tool that implements a multi-objective test prioritization technique based on the information related to the code and requirements coverage, as well as the execution cost of each test case. To establish users' and system requirements coverage, the MOTCP uses Latent Semantic Indexing to recover traceability links among application source code and requirements specifications. The test case ordering is then obtained by applying a non-dominated sorting genetic algorithm.

Distributed software systems are characterized by increasing autonomy. They often have the capability to sense the environment and react to it, discover the presence of other systems and take advantage of their services, adapt and re-configure themselves in accordance with the internal as well as the global state. Testing this kind of systems is challenging, and systematic and automated approaches are still missing. We propose a novel evolutionary testing framework for autonomous distributed systems. In this framework, test cases are continuously generated and executed. Our current implementation of the framework provides two techniques for the automated, continuous generation of test cases: (1) random; (2) evolutionary-mutation. Preliminary experimental results, obtained on a case study, are encouraging and indicate that evolutionary testing can complement effectively the manual one.

Traditional approaches to dynamic system analysis and metrics measurement are based on system code (both source, intermediate and executable code) instrumentation or need ad hoc support by the run-time environment. In these contexts, the measurement process is tricky, invasive and the results could be affected by the process itself making the data not germane. Moreover, the tool based on these approaches are difficult to customize, extend and often use since their properties are rooted at specific system details (e.g., special tools such as bytecode analyzers or virtual machine goodies such as the debugger interface) and require high efforts, skills and knowledges to be adapted. Notwithstanding its importance, software measurement is clearly a nonfunctional concern and should not impact on the software development and efficiency. Aspectoriented programming provides the mechanisms to deal with this kind of concern and to overcome the software measurement limitations. In this paper, we present a different approach to dynamic software measurements based on aspect-oriented programming and the corresponding support framework named AOPÂHiddenMetrics. The proposed approach makes the measurement process highly customizable and easy to use reducing its invasiveness and the dependency from the code knowledge.

Photovoltaics is amongst the most important technologies for renewable energy sources, and plays a key role in the development of a society with a smaller environmental footprint. Key parameters for solar cells are their energy conversion efficiency, their operating lifetime, and the cost of the energy obtained from a photovoltaic system compared to other sources. The optimization of these aspects involves the exploitation of new materials and development of novel solar cell concepts and designs. Both theoretical modeling and characterization of such devices require a comprehensive view including all scales from the atomic to the macroscopic and industrial scale. The different length scales of the electronic and optical degrees of freedoms specifically lead to an intrinsic need for multiscale simulation, which is accentuated in many advanced photovoltaics concepts including nanostructured regions. Therefore, multiscale modeling has found particular interest in the photovoltaics community, as a tool to advance the field beyond its current limits. In this article, we review the field of multiscale techniques applied to photovoltaics, and we discuss opportunities and remaining challenges. © T. Abu Hamed et al., published by EDP Sciences, 2018., European Cooperation in Science and Technology: MP1406, The authors are grateful for the financial support by the COST Action MP1406 “MultiscaleSolar.”

International audience; In this paper we address the problem of learning robust cross-domain representations for sketch-based image retrieval (SBIR). While most SBIR approaches focus on extracting low-and mid-level descriptors for direct feature matching, recent works have shown the benefit of learning coupled feature representations to describe data from two related sources. However, cross-domain representation learning methods are typically cast into non-convex minimization problems that are difficult to optimize, leading to unsatisfactory performance. Inspired by self-paced learning, a learning methodology designed to overcome convergence issues related to local optima by exploiting the samples in a meaningful order (i.e. easy to hard), we introduce the cross-paced partial curriculum learning (CPPCL) framework. Compared with existing self-paced learning methods which only consider a single modality and cannot deal with prior knowledge, CPPCL is specifically designed to assess the learning pace by jointly handling data from dual sources and modality-specific prior information provided in the form of partial curricula. Additionally, thanks to the learned dictionaries, we demonstrate that the proposed CPPCL embeds robust coupled representations for SBIR. Our approach is extensively evaluated on four publicly available datasets (i.e. CUFS, Flickr15K, QueenMary SBIR and TU-Berlin Extension datasets), showing superior performance over competing SBIR methods.

Palynological studies are considered as an excellent tool for palaeoaltitude estimates as sedimentary basins receive abundant pollen grains from surrounding uplands, especially through riverine detritic inputs. Here, we provide new evidence for estimating the palaeoaltitude of the Northern and Central Apennines (Italy) during the late Neogene (from the late Tortonian to the early Zandean), based on vegetation data derived from pollen analysis of five sections (Trave, Legnagnone, Cava Li Monti, Maccarone and Stirone). The past vegetation changes reflect shifts in the vegetation belts on the nearby massifs. The climate reconstruction realised on these sites, at sea-level, suggests that mean annual temperatures were higher than today at the end of the Neogene. Therefore, the different vegetation belts occurred at higher elevation than at present to compensate for the higher temperatures. The results obtained in our study suggest that the Northern and Central Apennines were uplifted as early as the Tortonian attaining significant minimum altitudes around 1350 m up to 1500 m at the end of the Miocene in the Central part, a minimum altitude around 1900 m during the Messinian and around 2150 m during the early Zandean in the Northern part. (C) 2014 Elsevier B.V. All rights reserved.

Self-supervised learning models have been shown to learn rich visual representations without requiring human annotations. However, in many real-world scenarios, labels are partially available, motivating a recent line of work on semi-supervised methods inspired by self-supervised principles. In this paper, we propose a conceptually simple yet empirically powerful approach to turn clustering-based self-supervised methods such as SwAV or DINO into semi-supervised learners. More precisely, we introduce a multi-task framework merging a supervised objective using ground-truth labels and a self-supervised objective relying on clustering assignments with a single cross-entropy loss. This approach may be interpreted as imposing the cluster centroids to be class prototypes. Despite its simplicity, we provide empirical evidence that our approach is highly effective and achieves state-of-the-art performance on CIFAR100 and ImageNet., Comment: CVPR 2023 - Code available at https://github.com/pietroastolfi/suave-daino

A complex manifold X X is weakly complete if it admits a continuous plurisubharmonic exhaustion function ϕ \phi . The minimal kernels Σ X k , k ∈ [ 0 , ∞ ] \Sigma _X^k, k \in [0,\infty ] (the loci where all C k \mathcal {C}^k plurisubharmonic exhaustion functions fail to be strictly plurisubharmonic), introduced by Slodkowski-Tomassini, and the Levi currents, introduced by Sibony, are both concepts aimed at measuring how far X X is from being Stein. We compare these notions, prove that all Levi currents are supported by all the Σ X k \Sigma _X^k ’s, and give sufficient conditions for points in Σ X k \Sigma _X^k to be in the support of some Levi current. When X X is a surface and ϕ \phi can be chosen analytic, building on previous work by the second author, Slodkowski, and Tomassini, we prove the existence of a Levi current precisely supported on Σ X ∞ \Sigma _X^\infty , and give a classification of Levi currents on X X . In particular, unless X X is a modification of a Stein space, every point in X X is in the support of some Levi current.

International audience; Contagion processes on networks, including disease spreading, information diffusion, or social behaviors propagation, can be modeled as simple contagion, i.e., as a contagion process involving one connection at a time, or as complex contagion, in which multiple interactions are needed for a contagion event. Empirical data on spreading processes, however, even when available, do not easily allow us to uncover which of these underlying contagion mechanisms is at work. We propose a strategy to discriminate between these mechanisms upon the observation of a single instance of a spreading process. The strategy is based on the observation of the order in which network nodes are infected, and on its correlations with their local topology: these correlations differ between processes of simple contagion, processes involving threshold mechanisms, and processes driven by group interactions (i.e., by “higher-order” mechanisms). Our results improve our understanding of contagion processes and provide a method using only limited information to distinguish between several possible contagion mechanisms.

Despite significant advances, the performance of state-of-the-art continual learning approaches hinges on the unrealistic scenario of fully labeled data. In this paper, we tackle this challenge and propose an approach for continual semi-supervised learning--a setting where not all the data samples are labeled. A primary issue in this scenario is the model forgetting representations of unlabeled data and overfitting the labeled samples. We leverage the power of nearest-neighbor classifiers to nonlinearly partition the feature space and flexibly model the underlying data distribution thanks to its non-parametric nature. This enables the model to learn a strong representation for the current task, and distill relevant information from previous tasks. We perform a thorough experimental evaluation and show that our method outperforms all the existing approaches by large margins, setting a solid state of the art on the continual semi-supervised learning paradigm. For example, on CIFAR-100 we surpass several others even when using at least 30 times less supervision (0.8% vs. 25% of annotations). Finally, our method works well on both low and high resolution images and scales seamlessly to more complex datasets such as ImageNet-100. The code is publicly available on https://github.com/kangzhiq/NNCSL, 13 pages

International audience; This work describes the evaluations of two approaches, Lexical Matching and Sense Similarity, for word sense alignment between MultiWordNet and a lexicographic dictionary, Senso Comune De Mauro, when having few sense descriptions (MultiWordNet) and no structure over senses (Senso Comune De Mauro). The results obtained from the merging of the two approaches are satisfying, with F1 values of 0.47 for verbs and 0.64 for nouns.

The Tiber Valley Project aims to create a series of digital applications for 3D reconstructing, visualization and real time browsing of the ancient and current Tiber Valley landscape (particularly for the Villa dei Volusii and Lucus Feroniae areas), in four different historical phases. In this perspective, the first problem to face is the need for a valid methodology for ancient landscape ecosystem reconstruction, before dealing with monuments and building. On the basis of an intense multi-disciplinary discussion and the previous VH Lab experience in this field, in this article we are presenting a scheme for a standardized reconstruction procedure, where the landscape is built using all available sources and elevation data obtained by a photogrammetry process on historical pictures. Ecosystem areas are then calculated through GIS elaboration in GRASS-GIS environment, through a procedure which may be shared for any situation of historical landscape reconstruction, allowing the matching and the mathematical processing of geographical data aimed to the definition of different ecological areas (both in terms of natural vegetation and cultivated lands). Maps are then created to be imported in procedural landscape generation engines: the last part of the paper focuses on the lack of effective open source software in this field, and a possible proposal implementation in this sense, 1. Il progetto " Valle del Tevere " Il presente lavoro illustra la sperimentazione di una metodologia per la ricostruzione del paesaggio storico, con particolare riguardo agli aspetti di definizione delle diverse presenze vegetazionali e antropiche del territorio in determinate epoche del passato. Il caso di studio è rappresentato dal lavoro che l'Istituto per le Tecnologie Applicate ai Beni Culturali (ITABC) del CNR sta svolgendo nell'ambito del Programma " Arcus " , finalizzato alla creazione di un sistema integrato di conoscenza, valorizzazione e comunicazione del paesaggio culturale della Valle del Tevere (contesti archeologici, storico-arti-stici, naturalistici, antropici), in particolare dell'area compresa tra il Monte Soratte e Fiano Romano, in direzione N-S, e il tracciato della via Flaminia antica e Palombara Sabina, in direzione E-O. Il progetto prevede una variegata serie di prodotti finali, quali: – un'installazione di realtà virtuale ludico-educativa, caratterizzata da un sistema di natural interaction (interazione attraverso i movimenti del corpo), localizzata all'interno di un museo di Roma, quale porta privilegiata di accesso e promozione del territorio della Valle del Tevere; – una guida multimediale alla Villa dei Volusii e al sito di Lucus Feroniae, da fruire sia su mobile che presso il Museo di Lucus Feroniae; – una guida alla Riserva del Tevere-Farfa, per smartphone o tablet, da fruire durante la visita all'area naturalistica; – un'installazione multimediale-filmica dedicata alla Riserva naturale del Teve-re-Farfa e destinata alle scuole, da fruire nel Museo del Fiume di Nazzano; – un sito web sul paesaggio culturale, basato su un sistema informativo geo-grafico in 3D e dedicato al pubblico di turisti, studiosi, operatori, scuole. Il lavoro comporta una ricostruzione fotorealistica del paesaggio natu-rale e antropico rispetto alle seguenti fasi cronologiche: – fase preistorica (ricostruzione della storia geologica dell'area, formazione della Valle del Tevere e dell'alveo fluviale, tra 3 milioni di anni fa e 12.000 anni fa); – fase pre-romana (focalizzata sull'Età del Ferro e sul periodo Orientaliz-zante); – fase romana (focalizzata, in particolare, sul periodo augusteo); – fase medievale (focalizzata, in particolare, sul XII secolo); – fase contemporanea.

International audience; Modelling languages and frameworks have been the key technology for advancing Model-Driven Engineering (MDE) methods and tools. Many industrial and research tools have been realised and are used across many domains. Hence, we think it is the right time to define what should be the future of modelling technologies, especially the requirements for the next generation of modelling frameworks and languages. In January 2020, the Second Winter Modelling Meeting (WMM2020) was held in San Vigilio di Marebbe (Italy), focusing on the analysis of the state of research, state of practice, and state of the art in MDE. The event brought together experts from industry, academia, and the open-source community to assess (i) what had changed in research on modelling in the last ten years, (ii) which problems are still unsolved, and (iii) which new challenges have arisen. This article presents a set of success stories and driving success factors of modelling and MDE, as well as a set of challenges and corresponding research directions that emerged from the synthesis of the results of our analysis.

International audience; No abstract available

A fundamental and challenging problem in deep learning is catastrophic forgetting, i.e. the tendency of neural networks to fail to preserve the knowledge acquired from old tasks when learning new tasks. This problem has been widely investigated in the research community and several Incremental Learning (IL) approaches have been proposed in the past years. While earlier works in computer vision have mostly focused on image classification and object detection, more recently some IL approaches for semantic segmentation have been introduced. These previous works showed that, despite its simplicity, knowledge distillation can be effectively employed to alleviate catastrophic forgetting. In this paper, we follow this research direction and, inspired by recent literature on contrastive learning, we propose a novel distillation framework, Uncertainty-aware Contrastive Distillation (\method). In a nutshell, \method~is operated by introducing a novel distillation loss that takes into account all the images in a mini-batch, enforcing similarity between features associated to all the pixels from the same classes, and pulling apart those corresponding to pixels from different classes. In order to mitigate catastrophic forgetting, we contrast features of the new model with features extracted by a frozen model learned at the previous incremental step. Our experimental results demonstrate the advantage of the proposed distillation technique, which can be used in synergy with previous IL approaches, and leads to state-of-art performance on three commonly adopted benchmarks for incremental semantic segmentation. The code is available at \url{https://github.com/ygjwd12345/UCD}., Comment: TPAMI

International audience; Microservices are an emerging architectural style advocating for small loosely-coupled services in order to maximize scalability and adaptability. In order to help IT personnel, adaptability can be put (completely or partially) under the responsibility of the system using autonomic techniques, e.g., underpinned by a MAPE-K control loop. This paper discusses possible trade-offs, challenges and support techniques for software architects involved in building autonomic microservicebased systems.

Although fully autonomous mapping methods are becoming more and more common and reliable, still the human operator is regularly employed in many 3D surveying missions. In a number of underwater applications, divers or pilots of remotely operated vehicles (ROVs) are still considered irreplaceable, and tools for real-time visualization of the mapped scene are essential to support and maximize the navigation and surveying efforts. For underwater exploration, image mosaicing has proved to be a valid and effective approach to visualize large mapped areas, often employed in conjunction with autonomous underwater vehicles (AUVs) and ROVs. In this work, we propose the use of a modified image mosaicing algorithm that coupled with image-based real-time navigation and mapping algorithms provides two visual navigation aids. The first is a classic image mosaic, where the recorded and processed images are incrementally added, named 2D sequential image mosaicing (2DSIM). The second one geometrically transform the images so that they are projected as planar point clouds in the 3D space providing an incremental point cloud mosaicing, named 3D sequential image plane projection (3DSIP). In the paper, the implemented procedure is detailed, and experiments in different underwater scenarios presented and discussed. Technical considerations about computational efforts, frame rate capabilities and scalability to different and more compact architectures (i.e. embedded systems) is also provided.

International audience; This preface for the proceedings of the first international workshop on Modular Knowledge (MK2022) presents the content of the workshop and reports on the results of the interactive sessions that took place during the event. The dramatic increase in the amount of open and linked data and the increasing semantification of such data make clear that knowledge is not monolithic, static or uniform. This requires a renewed push for dealing with heterogeneous and distributed knowledge as a constellation of modules. Each module stores a portion of knowledge about one particular subdomain, described in a specific schema, and valid under a set of circumstances. In such a scenario we need well-founded conceptual approaches and practical techniques for modular knowledge management, for example, to recognize relevant partitions of a monolithic knowledge source, but also to define a modularized vision of the domain qualifying the knowledge with a given situation or agent, integrating heterogeneous modules of knowledge, including knowledge represented in sub-symbolic models. The discussion of such modularity notions and techniques, their development and exploitation are the focus of the proposed Modular Knowledge workshop. The Modular Knowledge workshop combines the efforts of previous experiences (like WoMO, ARCOE-Logic and WOMoCoE workshops) into an interdisciplinary venue for discussing and developing solutions for modularity of knowledge. The workshop series aims at covering the use of various approaches (ranging from rich semantic representations, like Knowledge Graphs and formal ontology, to simpler schemas, like RDF and database schemas) for representing knowledge, its context, its evolution, and for making it accessible to automatic reasoning and knowledge management tasks. The Modular Knowledge workshop covers logic-based languages as well as subsymbolic and numerical representations.

Subtitles appear on screen as short pieces of text, segmented based on formal constraints (length) and syntactic/semantic criteria. Subtitle segmentation can be evaluated with sequence segmentation metrics against a human reference. However, standard segmentation metrics cannot be applied when systems generate outputs different than the reference, e.g. with end-to-end subtitling systems. In this paper, we study ways to conduct reference-based evaluations of segmentation accuracy irrespective of the textual content. We first conduct a systematic analysis of existing metrics for evaluating subtitle segmentation. We then introduce $Sigma$, a new Subtitle Segmentation Score derived from an approximate upper-bound of BLEU on segmentation boundaries, which allows us to disentangle the effect of good segmentation from text quality. To compare $Sigma$ with existing metrics, we further propose a boundary projection method from imperfect hypotheses to the true reference. Results show that all metrics are able to reward high quality output but for similar outputs system ranking depends on each metric's sensitivity to error type. Our thorough analyses suggest $Sigma$ is a promising segmentation candidate but its reliability over other segmentation metrics remains to be validated through correlations with human judgements., Comment: Accepted at LREC 2022

International audience; Over the past years, semantic segmentation, as many other tasks in computer vision, benefited from the progress in deep neural networks, resulting in significantly improved performance. However, deep architectures trained with gradient-based techniques suffer from catastrophic forgetting, which is the tendency to forget previously learned knowledge while learning new tasks. Aiming at devising strategies to counteract this effect, incremental learning approaches have gained popularity over the past years. However, the first incremental learning methods for semantic segmentation appeared only recently. While effective, these approaches do not account for a crucial aspect in pixel-level dense prediction problems, i.e. the role of attention mechanisms. To fill this gap, in this paper we introduce a novel attentive feature distillation approach to mitigate catastrophic forgetting while accounting for semantic spatial- and channel-level dependencies. Furthermore, we propose a {continual attentive fusion} structure, which takes advantage of the attention learned from the new and the old tasks while learning features for the new task. Finally, we also introduce a novel strategy to account for the background class in the distillation loss, thus preventing biased predictions. We demonstrate the effectiveness of our approach with an extensive evaluation on Pascal-VOC 2012 and ADE20K, setting a new state of the art.

International audience; The proceedings of the first international workshop on Modular Knowledge (MK2022).

International audience; Self-supervised models have been shown to produce comparable or better visual representations than their supervised counterparts when trained offline on unlabeled data at scale. However, their efficacy is catastrophically reduced in a Continual Learning (CL) scenario where data is presented to the model sequentially. In this paper, we show that self-supervised loss functions can be seamlessly converted into distillation mechanisms for CL by adding a predictor network that maps the current state of the representations to their past state. This enables us to devise a framework for Continual self-supervised visual representation Learning that (i) significantly improves the quality of the learned representations, (ii) is compatible with several state-of-the-art self-supervised objectives, and (iii) needs little to no hyperparameter tuning. We demonstrate the effectiveness of our approach empirically by training six popular self-supervised models in various CL settings.

In this paper we consider the problem of orchestrating video analytics applications over an edge computing infrastructure. Video analytics applications have been traditionally associated to the processing of video streams generated by fixed video cameras. Nowadays, however, the availability of mobile video cameras has become pervasive. We argue that to take advantage of the presence of mobile video cameras-and their informative content-it may be necessary to refactor the edge orchestration logic. We propose a new solution that splits the problem into two connected actions: 1) Placement of processing functions in the infrastructure and 2) Admission of most informative cameras based on their field of view. We hence describe a possible scheme for joint video stream admission and orchestration. Finally, preliminary numerical results are presented, demonstrating that separating the two logic components can improve coverage while reducing the cost of deployment. CCS CONCEPTS • Networks → Network control algorithms; • Information systems → Multimedia streaming.

In this paper, we are interested in audio-visual speech separation given a single-channel audio recording as well as visual information (lips movements) associated with each speaker. We propose an unsupervised technique based on audio-visual generative modeling of clean speech. More specifically, during training, a latent variable generative model is learned from clean speech spectrograms using a variational auto-encoder (VAE). To better utilize the visual information, the posteriors of the latent variables are inferred from mixed speech (instead of clean speech) as well as the visual data. The visual modality also serves as a prior for latent variables, through a visual network. At test time, the learned generative model (both for speaker-independent and speaker-dependent scenarios) is combined with an unsupervised non-negative matrix factorization (NMF) variance model for background noise. All the latent variables and noise parameters are then estimated by a Monte Carlo expectation-maximization algorithm. Our experiments show that the proposed unsupervised VAE-based method yields better separation performance than NMF-based approaches as well as a supervised deep learning-based technique., Comment: Accepted to the 31st IEEE International Workshop on Machine Learning for Signal Processing (MLSP), Oct. 25-28, 2021, Gold Coast, Queensland, Australia

This paper introduces Click to Move (C2M), a novel framework for video generation where the user can control the motion of the synthesized video through mouse clicks specifying simple object trajectories of the key objects in the scene. Our model receives as input an initial frame, its corresponding segmentation map and the sparse motion vectors encoding the input provided by the user. It outputs a plausible video sequence starting from the given frame and with a motion that is consistent with user input. Notably, our proposed deep architecture incorporates a Graph Convolution Network (GCN) modelling the movements of all the objects in the scene in a holistic manner and effectively combining the sparse user motion information and image features. Experimental results show that C2M outperforms existing methods on two publicly available datasets, thus demonstrating the effectiveness of our GCN framework at modelling object interactions. The source code is publicly available at https://github.com/PierfrancescoArdino/C2M., Comment: Accepted by International Conference on Computer Vision (ICCV 2021)

Context: Feedback from end users of software applications is a valuable resource in understanding what users request, what they value, and what they dislike. Information derived from user-feedback can support software evolution activities, such as requirements prioritization. User-feedback analysis is still mostly performed manually by practitioners, despite growing research in automated analysis. Objective: We address two issues in automated user-feedback analysis: (i) most of the existing automated analysis approaches that exploit linguistic analysis assume that the vocabulary adopted by users (when expressing feedback) and developers (when formulating requirements) are the same; and (ii) user-feedback analysis techniques are usually experimentally evaluated only on some user-feedback dataset, not involving assessment by potential software developers. Method: We propose an approach, ReFeed , that computes, for each requirement, the set of related user-feedback, and from such user-feedback extracts quantifiable properties which are relevant for prioritizing the requirement. The extracted properties are propagated to the related requirements, based on which ranks are computed for each requirement. ReFeed relies on domain knowledge, in the form of an ontology, helping mitigate the gap in the vocabulary of end users and developers. The effectiveness of ReFeed is evaluated on a realistic requirements prioritization scenario in two experiments involving graduate students from two different universities. Results: ReFeed is able to synthesize reasonable priorities for a given set of requirements based on properties derived from user-feedback. The implementation of ReFeed and related resources are publicly available. Conclusion: The results from our studies are encouraging in that using only three properties of user-feedback, ReFeed is able to prioritize requirements with reasonable accuracy. Such automatically determined prioritization could serve as a good starting point for requirements experts involved in the task of prioritizing requirements Future studies could explore additional user-feedback properties to improve the effectiveness of computed priorities.

Purpose: Old (>64 years) and very old (>79 years) intensive care patients with sepsis have a high mortality. In the very old, the value of critical care has been questioned. We aimed to compare the mortality, rates of organ support, and the length of stay in old vs. very old patients with sepsis and septic shock in intensive care.Methods: This analysis included 9,385 patients, from the multi-center eICU Collaborative Research Database, with sepsis; 6184 were old (aged 65–79 years), and 3,201 were very old patients (aged 80 years and older). A multi-level logistic regression analysis was used to fit three sequential regression models for the binary primary outcome of ICU mortality. A sensitivity analysis in septic shock patients (n = 1054) was also conducted.Results: In the very old patients, the median length of stay was shorter (50 ± 67 vs. 56 ± 72 h; p < 0.001), and the rate of a prolonged ICU stay was lower (>168 h; 9 vs. 12%; p < 0.001) than the old patients. The mortality from sepsis was higher in very old patients (13 vs. 11%; p = 0.005), and after multi-variable adjustment being very old was associated with higher odds for ICU mortality (aOR 1.32, 95% CI 1.09–1.59; p = 0.004). In patients with septic shock, mortality was also higher in the very old patients (38 vs. 36%; aOR 1.50, 95% CI 1.10–2.06; p = 0.01).Conclusion: Very old ICU-patients suffer from a slightly higher ICU mortality compared with old ICU-patients. However, despite the statistically significant differences in mortality, the clinical relevance of such minor differences seems to be negligible.

International audience; In both the industrial and scientific fields, the need for very high-resolution cartographic data is constantly increasing. With the aging of offshore subsea assets, it is very important to plan and maintain the longevity of structures, equipment, and systems. Inspection, maintenance, and repair (IMR) of subsea structures are key components of an overall integrity management system that aims to reduce the risk of failure and extend the life of installations. The acquisition of very detailed data during the inspection phase is a technological challenge, especially since offshore installations are sometimes deployed in extreme conditions (e.g., depth, hydrodynamics, visibility). After a review of high resolution mapping techniques for underwater environment, this article will focus on optical sensors that can satisfy the requirements of the offshore industry by assessing their relevance and degree of maturity. These requirements concern the resolution and accuracy but also cost, ease of implementation, and qualification. With the evolution of embedded computing resources, in-vehicle optical survey solutions are becoming increasingly important in the landscape of large-scale mapping solutions and more and more off-the-shelf systems are now available. The issues raised in this review are mainly related to the qualification of the results produced by optical systems and their limitations to cover all the needs expressed by the oil and gas industry field. Interesting qualification works of these solutions are presented in this paper as well as the use of online processing tools such as visual odometry or VSLAM to guide the data acquisition and pre-qualified survey. Finally, it seems interesting to combine acoustic and optical technologies in order to extend the field of application of these methods to low visibility conditions, which remains one of the main limiting factors in the generalization of the use of optical sensors in high resolution underwater cartography applications.

Photogrammetry needs known geometric elements to provide metric traceable measurements. These known elements can be a distance between two three-dimensional object points or two camera stations, or a combination of known coordinates and/or angles to solve the seven degrees of freedom that lead to rank deficiency of the normal-equation matrix. In this paper we present a novel approach for scaling and levelling to the local vertical direction an underwater photogrammetric survey. The developed methodology is based on a portable low-cost device designed and realized by the authors that uses depth measurements from a high resolution pressure sensor. The prototype consists of a data logger featuring a pressure sensor synchronized with a digital camera in its underwater pressure housing. The modular design, with optical communication and synchronization, provides great flexibility not requiring the camera housing to undergo any hardware modifications. The proposed methodology allows for a full 3D levelling transformation comprising two angles, a vertical translation and a scale factor and can work for surveying scenes extending horizontally, vertically or both. The paper presents the theoretical principles, an overview of the developed system together with preliminary calibration results. Tests in a lake and at sea are reported. An accuracy better than 1:5000 on the length measurement was achieved in calm water conditions.

International audience; With the experiment that we outline in this paper, we have had the ambition to pave the way for addressing the problem of supporting, enhancing and measuring collective AND informal learning, in particular serendipity. We want to support a new type of free navigation on Web resources (Documents, Topics, Events and Agents – human and artificial -) that is driven by the learner’s current needs and the preferences of the community of trust chosen by the learner, not by external actors. The experiment exploits the ViewpointS Web Application (VWA) prototype, that restructures a private version of a subset of the Web according to personalized choices in order to determine distances/proximities among resources. The process allows to enable, empower and measure the influence of members of the community of trust of the learner, on the learner’s choices when navigating in search of THE resources corresponding to THE immediate need, goal, strategy, wish. In the following, we will outline: 1. the rationale of our efforts and 2. the user’s reactions during the phase of -formal and informal- learning the functions and use of the prototypical software environment VWA, i.e.: a proof of concept for VWA.

Modern protocols for wireless sensor networks efficiently support multi-hop upward traffic from many sensors to a collection point, a key functionality enabling monitoring applications. However, the ever-evolving scenarios involving low-power wireless devices increasingly require support also for downward traffic, e.g., enabling a controller to issue actuation commands based on the monitored data. The IETF Routing Protocol for Low-power and Lossy Networks (RPL) is among the few tackling both traffic patterns. Unfortunately, its support for downward traffic is significantly unreliable and inefficient compared to its upward counterpart. We tackle this problem by extending RPL with mechanisms inspired by opposed, yet complementary, principles. At one extreme, we retain the route-based operation of RPL and devise techniques allowed by the standard but commonly neglected by popular implementations. At the other extreme, we rely on flooding as the main networking primitive. Inspired by these principles, we define three base mechanisms, integrate them in a popular RPL implementation, analyze their individual and combined performance, and elicit the resulting tradeoffs in scalability, reliability, and energy consumption. The evaluation relies on simulation, using both real-world topologies from a smart city scenario and synthetic grid ones, as well as on testbed experiments validating our findings from simulation. Results show that the combination of all three mechanisms into a novel protocol, T-RPL (i) yields high reliability, close to the one of flooding, (ii) with a low energy consumption, similar to route-based approaches, and (iii) improves remarkably the scalability of RPL with respect to downward traffic.

International audience; The coordination and control of hierarchically organized interacting agents is an important issue in many applications, e.g., harbor or warehouse automation. A formalism of agents as hierarchical input/output automata is proposed. A system of interacting agents is modeled as the parallel composition of their automata. We extend the usual parallel composition operation of I/O automata with a hierarchical composition operation for refining abstract tasks into lower-level subtasks. We provide an algorithm to synthesize hierarchically organized controllers to coordinate the agents' interactions in order to drive the system toward desired states. Our main contribution regards the formal definition, the representation, the theorems about its properties (i.e., the parallel and hierarchical composition are distributive operations), and the synthesis algorithm, proved to be complete and correct.

International audience; Collective learning has been advocated to be at the source for innovation, particularly as serendipity seems historically to have been the driving force not only behind innovation, but also behind scientific discovery and artistic creation. Informal learning is well known to represent the most significant learning effects in humans, far better than its complement: formal learning with predefined objectives. We have designed an approach—ViewpointS—based on a digital medium—the ViewpointS Web Application—that enables and enhances the processes for sharing knowledge within a group and is equipped with metrics aimed at assessing collective and informal learning. In this article, we introduce by giving a brief state of the art about collective and informal learning, then outline our approach and medium, and finally, present and exploit a real-life experiment aimed at evaluating the ViewpointS approach and metrics.