Search

Accurate estimation of ice thickness is essential for understanding and predicting the behaviors of glaciers, since the ice thickness provides valuable information about the glacier state and helps anticipate the evolution of whole glaciers systems. This latter knowledge is particularly important given the fact that glaciers act as natural reservoirs in the global water cycle. They are also indicators of the current and past state of global and local climate, as the properties and dynamics of glaciers reveal insights into the environmental conditions they have experienced.Current methods for estimating ice thickness face issues : field measurements such as the Ground Penetrating Range method can be costly and may not provide dense, continuous, renewable coverage. On the other hand, methods based on physical modeling can be computationally intensive and are dependent on assumptions about model parameters that may be unreliable in a poorly understood context. Results may thus be sensitive to the choice of prior information and prone to bias when working with limited or noisy data.Deep learning methods provide a promising solution for ice thickness prediction. One key advantage is their ability to handle large, multi-dimensional datasets and to learn directly from raw data without prior knowledge. Additionally, deep learning models are able to exploit non-linear relationships between datasets. Using such models also allows simultaneous training of other tasks, such as terrain classification to identify the presence of glaciers when it is not provided or outdated. In this study, we propose the use of such an approach relying on convolutional models based on VGGNet, ResNet and U-Net.Our goal is to obtain an accurate estimation of the glacier thickness distribution. We propose the use of neural networks in order to 1) be free from statistical/physical assumptions, 2) leverage deep relationships between observed data and physical parameters to be estimated, 3) overcome inaccuracies in collected data, and 4) accurately represent complex patterns such as non-linear thickness variations within the glacier. Additionally, it is important that these models should not be prone to common issues of deep learning such as overfitting and lack of explainability.We conduct our study on Alpine glaciers in Switzerland. The input data for our neural network models includes: 1) average ice velocity fields calculated from correlation of Sentinel-2 images with a resolution of 50 metres, and 2) altitudes and slopes derived from the Swiss digital elevation model with a resolution of 10 metres. To verify the accuracy of the predicted ice thickness values, we use ground truth data obtained from GPR surveys conducted in profile form, from 2012 to 2021.In addition to estimating the ice thickness, we also perform direct classification of glaciers vs. non-glacier areas. Results demonstrate the feasibility of quickly training a neural network model with limited training data and producing stable, high-quality ice thickness estimates for different glaciers in the study region.

Sur un problème d'analyse de rayonnement gamma à partir d'observations depuis des télescopes à imagerie Cherenkov, ce chapitre présente un modèle de réseau de neurones profond multitâche. Celui-ci permet la reconstruction des paramètres des rayonnements observés. Nous démontrons l’intérêt de l'approche multitâche. Nous montrons également que cette architecture obtient de meilleures performances qu’une méthode d’analyse standard largement utilisée pour cette problématique.

Las discusiones sobre el centro civico de la nueva capital de Brasil remontan a un arco que empieza en las comisiones tecnicas despues del reporte Belcher en 1954 hasta la Plaza de los Tres Poderes disenada por Oscar Niemeyer y Lucio Costa en 1957. Permanece oscuro el nexo entre la arquitectura de la plaza civica y sus expectaciones en el medio politico, la opinion publica y en los medios tecnicos. Destacase la ausencia de textos que detallen el confronto entre el recien juramentado presidente Juscelino Kubitschek y el Mariscal Jose Pessoa, incumbente del planeamiento de la ciudad. La revision bibliografica de los periodicos locales y la documentacion primaria demuestran los meandros de una disputa que mejor cualifica la subscricion de la obra de Oscar Niemeyer a las pretensiones de poder cultural intentado por Kubitschek mientras se detallan como el cuerpo tecnico de Pessoa y, posteriormente, Niemeyer y Costa desarrollaran distintas arquitecturas para la plaza de los palacios decisorios del pais.

Ocean surface monitoring, emphasizing oil slick detection, has become essential due to its importance for oil exploration and ecosystem risk prevention. Automation is now mandatory since the manual annotation process of oil by photo-interpreters is time-consuming and cannot process the data collected continuously by the available spaceborne sensors. Studies on automatic detection methods mainly focus on Synthetic Aperture Radar (SAR) data exclusively to detect anthropogenic (spills) or natural (seeps) oil slicks, all using limited datasets. The main goal is to maximize the detection of oil slicks of both natures while being robust to other phenomena that generate false alarms, called “lookalikes”. To this end, this paper presents the automation of offshore oil slick detection on an extensive database of real and recent oil slick monitoring scenarios, including both types of slicks. It relies on slick annotations performed by expert photo-interpreters on Sentinel-1 SAR data over four years and three areas worldwide. In addition, contextual data such as wind estimates and infrastructure positions are included in the database as they are relevant data for oil detection. The contributions of this paper are: (i) A comparative study of deep learning approaches using SAR data. A semantic and instance segmentation analysis via FC-DenseNet and Mask R-CNN, respectively. (ii) A proposal for Fuse-FC-DenseNet, an extension of FC-DenseNet that fuses heterogeneous SAR and wind speed data for enhanced oil slick segmentation. (iii) An improved set of evaluation metrics dedicated to the task that considers contextual information. (iv) A visual explanation of deep learning predictions based on the SHapley Additive exPlanation (SHAP) method adapted to semantic segmentation. The proposed approach yields a detection performance of up to 94% of good detection with a false alarm reduction ranging from 14% to 34% compared to mono-modal models. These results provide new solutions to improve the detection of natural and anthropogenic oil slicks by providing tools that allow photo-interpreters to work more efficiently on a wide range of marine surfaces to be monitored worldwide. Such a tool will accelerate the oil slick detection task to keep up with the continuous sensor acquisition. This upstream work will allow us to study its possible integration into an industrial production pipeline. In addition, a prediction explanation is proposed, which can be integrated as a step to identify the appropriate methodology for presenting the predictions to the experts and understanding the obtained predictions and their sensitivity to contextual information. Thus it helps them to optimize their way of working.

Scholars of the reign of Ivan the Terrible have only a limited number of sources to understand Ivan’s purges during the Oprichnina. One of the most important and most commonly used testimonies for Ivan’s period of terror, the account of the Livonian captives Johann Taube and Elert Kruse, is for the first time made accessible for the English-speaking public in a critical edition that covers the major historiographical issues raised in the account and, when possible, explores the reasons behind the execution of specific individuals.

Sex differences in vertebrate spatial abilities are typically interpreted under the adaptive specialization hypothesis, which posits that male reproductive success is linked to larger home ranges and better navigational skills. The androgen spillover hypothesis counters that enhanced male spatial performance may be a byproduct of higher androgen levels. Animal groups that include species where females are expected to outperform males based on life-history traits are key for disentangling these hypotheses. We investigated the association between sex differences in reproductive strategies, spatial behavior, and androgen levels in three species of poison frogs. We tracked individuals in natural environments to show that contrasting parental sex roles shape sex differences in space use, where the sex performing parental duties shows wider-ranging movements. We then translocated frogs from their home areas to test their navigational performance and found that the caring sex outperformed the non-caring sex only in one out of three species. In addition, males across species displayed more explorative behavior than females and androgen levels correlated with explorative behavior and homing accuracy. Overall, we reveal that poison frog reproductive strategies shape movement patterns but not necessarily navigational performance. Together this work suggests that prevailing adaptive hypotheses provide an incomplete explanation of sex differences in spatial abilities.

Ocean surface monitoring, especially oil slick detection, has become mandatory due to its importance for oil exploration and risk prevention on ecosystems. For years, the detection task has been performed manually by photo-interpreters using Synthetic Aperture Radar (SAR) images with the help of contextual data such as wind. This tedious manual work cannot handle the increasing amount of data collected by the available sensors and thus requires automation. Literature reports conventional and semi-automated detection methods that generally focus either on oil slicks originating from anthropogenic (spills) or natural (seeps) sources on limited data collections. As an extension, this paper presents the automation of offshore oil slicks on an extensive database with both kinds of slicks. It builds upon the slick annotations of specialized photo-interpreters on Sentinel-1 SAR data for 6 years over 3 exploration and monitoring areas worldwide. All the considered SAR images and related annotation relate to real oil slick monitoring scenarios. Further, wind estimation is systematically computed to enrich the data collection. Paper contributions are the following : (i) a performance comparison of two deep learning approaches: semantic segmentation using FC-DenseNet and instance segmentation using Mask-RCNN. (ii) the introduction of meteorological information (wind speed) is deemed valuable for oil slick detection in the performance evaluation. The main results of this study show the effectiveness of slick detection by deep learning approaches, in particular FC-DenseNet, which captures more than 92% of oil instances in our test set. Furthermore, a strong correlation between model performances and contextual information such as slick size and wind speed is demonstrated in the performance evaluation. This work opens perspectives to design models that can fuse SAR and wind information to reduce the false alarm rate.

International audience; Deep Neural Networks (DNNs) are getting increasing attention to deal with Land Cover Classification (LCC) relying on Satellite Image Time Series (SITS). Though high performances can be achieved, the rationale of a prediction yielded by a DNN often remains unclear. An architecture expressing predictions with respect to input channels is thus proposed in this paper. It relies on convolutional layers and an attention mechanism weighting the importance of each channel in the final classification decision. The correlation between channels is taken into account to set up shared kernels and lower model complexity. Experiments based on a Sentinel-2 SITS show promising results.; Les réseaux profonds de neurones sont de plus en plus plébiscités, à fortiori, en classification d'occupation des sols, à partir de séries temporelles d'images satellites. Malgré leurs performances, ces réseaux souffrent de manque d'explicabilité. Dans ce cadre, nous étudions différentes architectures de réseaux profonds convolutionnels qui extraient les caractéristiques spectro-temporelles des séries pour ensuite classifier chaque pixel. Afin d'expliquer ses décisions, nous y intégrons, de deux manières différentes, un module d'attention donnant la contribution de chaque caractéristique spectrale. Des expériences ont été menées sur des données Sentinel-2 de la Réunion. Les corrélations spectrales, inhérentes à ce type de séries, sont prises en compte en factorisant les convolutions sur des bandes fortement corrélées afin d'obtenir le meilleur compromis complexité/performance du modèle. Les résultats semblant prometteurs en termes d'explicabilité et de performances, des développements sont à l'étude pour étendre les attentions au niveau spectral et temporel (cartes de chaleur) et pour comprendre plus finement le fonctionnement des réseaux "multi-sorties" qui permettent d'entraîner le module d'attention.

A Igreja da Pampulha, feita em Belo Horizonte em 1943, por Oscar Niemeyer, e reconhecida como marco das vanguardas arquitetonicas brasileiras. Seu programa religioso colocava‑se como desafio para o arquiteto, pelo primado de simbolos e significados que remontavam ao passado colonial das catedrais e capelas valorizadas pelos intelectuais modernistas, como Drummond de Andrade, Lucio Costa, Mario de Andrade. Entre os anos 1920 e 1930, e intenso o debate intelectual acerca da heranca do barroco mineiro como portadora de uma forca de atualizacao do presente que sintetizaria modernidade e identidade nacional. As capelas e arquiteturas de autoria nao definida sao contrapostas as monumentais criacoes de artifices notaveis, como Aleijadinho. Niemeyer nao era alheio a este debate sobre como o passado colonial seria referenciado nas obras de vanguarda, preferindo na sua igreja construir correspondencias tipologicas com as capelas de Ouro Preto e arredores construidas no seculo XVIII, problematizando, na arquitetura, questoes da ordem do dia dos modernistas, no que tange a relacao com o passado no presente.

Melanomas commonly undergo a phenotype switch, from a proliferative to an invasive state. Such tumor cell plasticity contributes to immunotherapy resistance; however, the mechanisms are not completely understood and thus are therapeutically unexploited. Using melanoma mouse models, we demonstrated that blocking the MNK1/2-eIF4E axis inhibited melanoma phenotype switching and sensitized melanoma to anti–PD-1 immunotherapy. We showed that phospho-eIF4E–deficient murine melanomas expressed high levels of melanocytic antigens, with similar results verified in patient melanomas. Mechanistically, we identified phospho-eIF4E–mediated translational control of NGFR, a critical effector of phenotype switching. Genetic ablation of phospho-eIF4E reprogrammed the immunosuppressive microenvironment, exemplified by lowered production of inflammatory factors, decreased PD-L1 expression on dendritic cells and myeloid-derived suppressor cells, and increased CD8(+) T cell infiltrates. Finally, dual blockade of the MNK1/2-eIF4E axis and the PD-1/PD-L1 immune checkpoint demonstrated efficacy in multiple melanoma models regardless of their genomic classification. An increase in the presence of intratumoral stem-like TCF1(+)PD-1(+)CD8(+) T cells, a characteristic essential for durable antitumor immunity, was detected in mice given a MNK1/2 inhibitor and anti–PD-1 therapy. Using MNK1/2 inhibitors to repress phospho-eIF4E thus offers a strategy to inhibit melanoma plasticity and improve response to anti–PD-1 immunotherapy.

As the coronavirus disease 2019 (COVID-19) pandemic second wave is emerging, it is of the upmost importance to screen the population immunity in order to keep track of infected individuals. Consequently, immunoassays for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) with high specificity and positive predictive values are needed to obtain an accurate epidemiological picture., As the coronavirus disease 2019 (COVID-19) pandemic second wave is emerging, it is of the upmost importance to screen the population immunity in order to keep track of infected individuals. Consequently, immunoassays for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) with high specificity and positive predictive values are needed to obtain an accurate epidemiological picture. As more data accumulate about the immune responses and the kinetics of neutralizing-antibody (nAb) production in SARS-CoV-2-infected individuals, new applications are forecast for serological assays such as nAb activity prediction in convalescent-phase plasma from recovered patients. This multicenter study, involving six hospital centers, determined the baseline clinical performances, reproducibility, and nAb level correlations of 10 commercially available immunoassays. In addition, three lateral-flow chromatography assays were evaluated, as these devices can be used in logistically challenged areas. All assays were evaluated using the same patient panels in duplicate, thus enabling accurate comparison of the tests. Seven immunoassays examined in this study were shown to have excellent specificity (98 to 100%) and good to excellent positive predictive values (82 to 100%) when used in a low (5%)-seroprevalence setting. We observed sensitivities as low as 74% and as high as 95% at ≥15 days after symptom onset. The determination of optimized cutoff values through receiver operating characteristic (ROC) curve analyses had a significant impact on the diagnostic resolution of several enzyme immunoassays by increasing the sensitivity significantly without a large trade-off in specificity. We found that spike-based immunoassays seem to be better correlates of nAb activity. Finally, the results reported here will add to the general knowledge of the interlaboratory reproducibility of clinical performance parameters of immunoassays and provide new evidence about nAb activity prediction.

Segmentation is a fundamental problem but not the ultimate goal, it is a stepping stone to higher level application problems. It consists in associating each of the low-level image pixels to the class they locally represent. This task completes image analysis tasks such as visual scene classification and instance level object detection. It enables high level applications in a variety of domains, from images and video indexing to autonomous vehicle driving and medical image analysis. Recently, deep learning approaches have pushed image segmentation and object instance segmentation in a new era with impressive performance levels. However, several challenges have to be faced to train those approaches in an effective way for each of the case studies, dealing with few training samples, specific data, strong target imbalance and so on. This chapter reviews the image segmentation task and recent advanced strategies to face those potential issues in a variety of application domains. Current challenges to address are highlighted for future research directions.

CUPID-Mo is a cryogenic detector array designed to search for neutrinoless double-beta decay (0νββ) of 100Mo. It uses 20 scintillating 100Mo-enriched Li2MoO4 bolometers instrumented with Ge light detectors to perform active suppression of α backgrounds, drastically reducing the expected background in the 0νββ signal region. As a result, pileup events and small detector instabilities that mimic normal signals become non-negligible potential backgrounds. These types of events can in principle be eliminated based on their signal shapes, which are different from those of regular bolometric pulses. We show that a purely data-driven principal component analysis based approach is able to filter out these anomalous events, without the aid of detector response simulations.

Melanomas commonly undergo a phenotype switch, from a proliferative to an invasive state. Melanoma plasticity exhibited as phenotype switching contributes to immunotherapy resistance, however the mechanisms are not completely understood and thus therapeutically unexploited. Here, using a transgenic melanoma mouse model, we demonstrated a critical role of the MNK1/2-eIF4E axis in melanoma plasticity and resistance to immunotherapy. We showed that phospho-eIF4E deficient murine melanomas express high levels of melanocytic antigens, with similar results verified in patient melanomas. Mechanistically, we identified that phospho-eIF4E controls the translation of NGFR, a critical effector of phenotype switching. In patients with melanoma, the expression of MKNK1, the kinase for eIF4E, positively correlated with markers of immune exhaustion. Genetic ablation of phospho-eIF4E reprogrammed the immunosuppressive microenvironment, exemplified by lowered production of inflammatory factors and increased CD8+ T cell infiltrates. Blocking phospho-eIF4E, using MNK1/2 inhibitors, offers a new strategy to inhibit melanoma plasticity and improve the survival response to anti-PD-1 immunotherapy.

Object recognition problem based on cameras has become one of the most important tasks in robotics. Nevertheless, for a robot equipped with a camera and evolving in uncontrolled environment, many factors can limit the quality of its recognition, such as lighting conditions, conflict situations, occlusion, and cameras orientations, etc. Under theses imperfections, a single robot camera can provide incomplete and unreliable information and is thus limited to deal with complex recognition situations, particularly in imprecise and uncertain environments. Under such circumstances, in this paper, an IoT (Internet of Things) multi-camera embedded 3D object recognition system, including a NAO humanoid robot and other IoT heterogeneous cameras (2D and 3D), is proposed. These cameras can be complementary, and their combination increases the recognition reliability of the single robot camera. The used IoT infrastructure allows the communication between the robot and different connected cameras using internet protocols. The proposed recognition system uses evidence theory to fuse the data issued from these IoT sources for making the recognition decision. This paper aims at demonstrating the advantages of coupling the IoT technology with the evidence theory concepts for improving the 3D object recognition for the NAO robot. Through the proposed application, our method has been validated and its performance has been compared with a deep learning approach.

The CUPID-Mo experiment is searching for neutrinoless double beta decay in $^{100}$Mo, evaluating the technology of cryogenic scintillating Li$_{2}^{100}$MoO$_4$ detectors for CUPID (CUORE Upgrade with Particle ID). CUPID-Mo detectors feature background suppression using a dual-readout scheme with Li$_{2}$MoO$_4$ crystals complemented by Ge bolometers for light detection. The detection of both heat and scintillation light signals allows the efficient discrimination of $\alpha$ from $\gamma$&$\beta$ events. In this proceedings, we discuss results from the first 2 months of data taking in spring 2019. In addition to an excellent bolometric performance of 6.7$\,$keV (FWHM) at 2615$\,$keV and an $\alpha$ separation of better than 99.9\% for all detectors, we report on bulk radiopurity for Th and U. Finally, we interpret the accumulated physics data in terms of a limit of $T_{1/2}^{0\nu}\,> 3\times10^{23}\,$yr for $^{100}$Mo and discuss the sensitivity of CUPID-Mo until the expected end of physics data taking in early 2020., Comment: Proceedings for TAUP 2019, submitted to IOP Journal of Physics: Conference Series

Recently, a variety of approaches have been enriching the field of remote sensing (RS) image processing and analysis. Unfortunately, existing methods remain limited to the rich spatiospectral content of today’s large data sets. It would seem intriguing to resort to deep learning (DL)-based approaches at this stage with regard to their ability to offer accurate semantic interpretation of the data. However, the specificity introduced by the coexistence of spectral and spatial content in the RS data sets widens the scope of the challenges presented to adapt DL methods to these contexts. Therefore, the aim of this paper is first to explore the performance of DL architectures for the RS hyperspectral data set classification and second to introduce a new 3-D DL approach that enables a joint spectral and spatial information process. A set of 3-D schemes is proposed and evaluated. Experimental results based on well-known hyperspectral data sets demonstrate that the proposed method is able to achieve a better classification rate than state-of-the-art methods with lower computational costs.

International audience; Late Jurassic and Early Cretaceous charophyte assemblages from the northern part of the Aquitaine Basin in south-west France are reviewed here to understand their palaeoecological, palaeobiogeographical and biostratigraphic features. Three sites were studied: the Tithonian-lower Berriasian of Chassiron, and the Berriasian of Cherves-de-Cognac and Angeac-Charente. Abundant porocharaceans, less abundant clavatoraceans and scarce characeans recorded in Cherves-de-Cognac and Angeac-Charente indicate that brackish water environments were substituted by freshwater environments eastwards. The occurrence of Clavator grovesii var. grovesii and morphotypes intermediate with C. grovesii var. discordis in the same areas is significant from a biostratigraphic viewpoint, since these species belong to the Maillardii, Incrassatus and Nurrensis European charophyte biozones, representing the Berriasian. This observation refutes a previous dating of the Angeac-Charente site and highlights the absence of Hauterivian–Barremian records in northern Aquitaine, which is in contrast to the more complete Lower Cretaceous record in southern Aquitaine. These contrasting records could be due to differences in the available sedimentary space produced by the opening of the Bay of Biscay during the Barremian.

Two try-out campaigns of friction stir welding (FSW) were performed with different friction parameters to join S690QL high yield strength steel. The welds were investigated at macroscopic and microscopic scales using optical and electronic microscopy and microhardness mapping. Welds of the second campaign exhibit microstructures and mechanical properties in accordance with requirements for service use. Microtexture measurements were carried out in different zones of welds by electron backscattered diffraction (EBSD). It is shown that that texture of the bottom of the weld is similar to that of the base metal, suggesting a diffusion bonding mechanism. Finally, the mechanical properties (tensile strength, resilience, bending) were established on the most promising welds. It is shown that it is possible to weld this high yield strength steel using FSW process with satisfactory geometric, microstructural and mechanical properties.

International audience; Imaging atmospheric Cherenkov telescopes (IACT) data require an important analysis in order to reconstruct events and obtain a photon list. The state-of-the-art reconstruction is made of several steps including image analysis, features extraction and machine learning. Since the 2012 Ima-geNet breakthrough, deep learning advances have shown dramatic improvements in data analysis across a variety of fields. Convolutional neural networks look particularly suited to the task of analysing IACT camera images for event reconstruction as they provide a way to reconstruct photon list directly from raw images, skipping the pre-processing steps. Moreover, despite demanding important computing resources to be trained and optimised, neural networks show very good performances during execution, making them viable for real-time analysis for the future generation of IACT. Here we present GammaLearn , a python framework providing the tools and environment to easily train neural networks on IACT data. Relying on PyTorch, it allows the use of indexed convolution on images with non-cartesian pixel lattices predominant in IACT for the low-level operations and offers a simple configuration file-based workflow, producing the trained model, training estimators as well as higher level results. The proposed framework is modular and straightforward to customize by end users. It has been tested and validated on the analysis of the Cherenkov Telescope Array simulated data.

International audience; Five lots of Foraminifera, belonging to the Marin Defrance (1758–1850) collection, were discovered in the collection of the Musée des Sciences in Laval (Mayenne, France). The original material used by Defrance to describe Alveolina boscii (Defrance in Bronn, 1825) and Fabularia discolites Defrance in Bronn, 1825 was identified among them, allowing the designation of a lectotype and paralectotypes for those two species. The systematic description of A. boscii and F. discolites is reviewed. The presence of these specimens in the museum collections is still puzzling, since the complete Defrance's collection was thought to have been lost during World War II.

International audience; The present paper introduces convolution and pooling operators for indexed images. These operators can be used on images that do not provide Cartesian grids of pixels, as long as a list of neighbor's indices can be provided for each pixel. They are foreseen being useful for convolutional neural networks (CNN) applied to special sensors, especially in science, without requiring image pre-processing. The present work explains the method and its implementation in the Pytorch framework and shows an application of the indexed kernels to the classification task of images with hexagonal lattices using CNN. The obtained results show that the method gives the same performances as the standard convolution kernels. Indexed convolution thus makes deep neural network frameworks more general and capable of addressing unconventional image lattices. The current implementation, as well as code to reproduce the experiments described in this paper are made available as open-source resources on the repository www.github.com/IndexedConv.

International audience; Cretaceous fossil inclusions were recently discovered in silica-rich nodules from the diachronic alteritic deposits of the Moragne hill, Charente-Maritime, western France. Nodules contain marine organisms such as sponge spicules, putative algae, echinoids (Periaster undulatus Agassiz and Desor) and rudists (Sphaerulites foliaceus Lamarck; hippuritids and radiolitids). These inclusions co-occur with fragmented pin-nae of ferns as well as diverse structures of conifers including leafy axes (Geinitzia sp.; Glenrosa carentonensis Moreau, Néraudeau, Tafforeau and Dépré) and isolated ovu-late scales or subtending bracts of cones. The palaeontological assemblage shows that the alterites from Moragne are condensed deposits resulting from the alteration of the Cenomanian to Turonian-Senonian regional series. Conifer leafy axes display diverse kinds of preservation: permineralization in silica, empty external casts of the gross morphology, and casts filled by ferruginous siltstone. Propagation phase-contrast X-ray synchrotron microtomography revealed exceptional preservation of some leafy axes up to the cell level. We support a multi-phase silicification including at least: (1) a precocious phase occurring during the sedimentation of the Upper Cretaceous deposits or during the earliest diagenesis, and that could be linked with the abundance of sponge spicules; then (2) a late phase occurring during Eocene-Oligocene period that corresponds to a hard climate event with strong and long paedogenitic alteration associated with an important phase of silicification (nodulation). The alterites from Moragne are an uncommon example of long-term time-averaged deposits caused by compac-tion and selective dissolution of matrix in the context of a late continental weathering.

International audience; In recent years, semantic segmentation has become one of the most active tasks of the computer vision field. Its goal is to group image pixels into semantically meaningful regions. Deep learning methods, in particular those who use convolutional neural network (CNN), have shown a big success for the semantic segmentation task. In this paper, we will introduce a semantic segmentation system using a reinforced fully convolutional densenet with multiscale kernel prediction method. Our main contribution is to build an encoder-decoder based architecture where we increase the width of dense block in the encoder part by conducting recurrent connections inside the dense block. The resulting network structure is called wider dense block where each dense block takes not only the output of the previous layer but also the initial input of the dense block. These recurrent structure emulates the human brain system and helps to strengthen the extraction of the target features. As a result, our network becomes deeper and wider with no additional parameters used because of weights sharing. Moreover, a multiscale convolutional layer has been conducted after the last dense block of the decoder part to perform model averaging over different spatial scales and to provide a more flexible method. This proposed method has been evaluated on two semantic segmentation benchmarks: CamVid and Cityscapes. Our method outperforms many recent works from the state of the art.

International audience; We compare a recent dehazing method based on deep learning, Dehazenet, with traditional state-of-the-art approaches , on benchmark data with reference. Dehazenet estimates the depth map from transmission factor on a single color image, which is used to inverse the Koschmieder model of imaging in the presence of haze. In this sense, the solution is still attached to the Koschmieder model. We demonstrate that the transmission is very well estimated by the network, but also that this method exhibits the same limitation than others due to the use of the same imaging model.

International audience; The Cherenkov Telescope Array (CTA) is the next generation of ground-based gamma-ray telescopes for gamma-ray astronomy. Two arrays will be deployed composed of 19 telescopes in the Northern hemisphere and 99 telescopes in the Southern hemisphere. Due to its very high sensitivity, CTA will record a colossal amount of data that represent a computing challenge to the reconstruction software. Moreover, the vast majority of triggered events come from protons that represent a background for gamma-ray astronomy. Deep learning developments in the last few years have shown tremendous improvements in the analysis of data in many domains. Thanks to the huge amount of simulated data and later of real data, produced by CTA, these algorithms look particularly adapted and very promising. Moreover, the trained neural networks show very good computing performances during execution. Here we present a first study of deep learning architectures applied to CTA simulated data to perform the reconstruction of the particles energy and incoming direction and the development of a specific framework, GammaLearn, to accomplish this task.

International audience; We compare a recent dehazing method based on deep learning , Dehazenet, with traditional state-of-the-art approach, on benchmark data with reference. Dehazenet estimates the depth map from a single color image, which is used to inverse the Koschmieder model of imaging in the presence of haze. In this sense, the solution is still attached to the Koschmieder model. We demonstrate that this method exhibits the same limitation than other inversions of this imaging model.

As a general rule, data analytics are now mandatory for companies. Scanned document analysis brings additional challenges introduced by paper damages and scanning quality.In an industrial context, this work focuses on the automatic understanding of sale receipts which enable access to essential and accurate consumption statistics. Given an image acquired with a smart-phone, the proposed work mainly focuses on the first steps of the full tool chain which aims at providing essential information such as the store brand, purchased products and related prices with the highest possible confidence. To get this high confidence level, even if scanning is not perfectly controlled, we propose a double check processing tool-chain using Deep Convolutional Neural Networks (DCNNs) on one hand and more classical image and text processings on another hand.The originality of this work relates in this double check processing and in the joint use of DCNNs for different applications and text analysis., International Conference on Image Processing Theory, Tools and Applications, Nov 2017, Montreal, Canada. 2017, http://www.ipta-conference.com/ipta17

International audience; The IRIM group is a consortium of French teams working on Multimedia Indexing and Retrieval. This paper describes its participation to the TRECVID 2017 instance search task.

International audience; This paper proposes to investigate the potential benefit of the use of low-level human vision behaviors in the context of high-level semantic concept detection. A large part of the current approaches relies on the Bag-of-Words (BoW) model, which has proven itself to be a good choice especially for object recognition in images. Its extension from static images to video sequences exhibits some new problems to cope with, mainly the way to use the temporal information related to the concepts to detect (swimming, drinking...). In this study, we propose to apply a human retina model to preprocess video sequences before constructing the State-Of-The-Art BoW analysis. This preprocessing, designed in a way that enhances relevant information, increases the performance by introducing robustness to traditional image and video problems, such as luminance variation, shadows, compression artifacts and noise. Additionally, we propose a new segmentation method which enables a selection of low-level spatio-temporal potential areas of interest from the visual scene, without slowing the computation as much as a high-level saliency model would. These approaches are evaluated on the TrecVid 2010 and 2011 Semantic Indexing Task datasets, containing from 130 to 346 high-level semantic concepts. We also experiment with various parameter settings to check their effect on performance.

International audience; The IRIM group is a consortium of French teams working on Multimedia Indexing and Retrieval. This paper describes its participation to the TRECVID 2016 instance search task.