Dissertations / Theses on the topic 'Détection de faux en profondeur'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 28 dissertations / theses for your research on the topic 'Détection de faux en profondeur.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Tak, Hemlata. "End-to-End Modeling for Speech Spoofing and Deepfake Detection." Electronic Thesis or Diss., Sorbonne université, 2023. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2023SORUS104.pdf.
Full textVoice biometric systems are being used in various applications for secure user authentication using automatic speaker verification technology. However, these systems are vulnerable to spoofing attacks, which have become even more challenging with recent advances in artificial intelligence algorithms. There is hence a need for more robust, and efficient detection techniques. This thesis proposes novel detection algorithms which are designed to perform reliably in the face of the highest-quality attacks. The first contribution is a non-linear ensemble of sub-band classifiers each of which uses a Gaussian mixture model. Competitive results show that models which learn sub-band specific discriminative information can substantially outperform models trained on full-band signals. Given that deep neural networks are more powerful and can perform both feature extraction and classification, the second contribution is a RawNet2 model. It is an end-to-end (E2E) model which learns features directly from raw waveform. The third contribution includes the first use of graph neural networks (GNNs) with an attention mechanism to model the complex relationship between spoofing cues present in spectral and temporal domains. We propose an E2E spectro-temporal graph attention network called RawGAT-ST. RawGAT-ST model is further extended to an integrated spectro-temporal graph attention network, named AASIST which exploits the relationship between heterogeneous spectral and temporal graphs. Finally, this thesis proposes a novel data augmentation technique called RawBoost and uses a self-supervised, pre-trained speech model as a front-end to improve generalisation in the wild conditions
Cord, Matthieu. "Analyse d'images aériennes haute résolution : détection et modélisation du bâti en zone urbaine." Cergy-Pontoise, 1998. http://biblioweb.u-cergy.fr/theses/98CERG0054.pdf.
Full textM'Saad, Soumaya. "Détection de changement de comportement de vie chez la personne âgée par images de profondeur." Thesis, Rennes 1, 2022. http://www.theses.fr/2022REN1S039.
Full textThe number of elderly people in the world is constantly increasing, hence the challenge of helping them to continue to live at home and ageing in good health. This PhD takes part in this public health issue and proposes the detection of the person behavior change based on the recording of activities in the home by low-cost depth sensors that guarantee anonymity and that operate autonomously day and night. After an initial study combining image classification by machine learning approaches, a method based on Resnet-18 deep neural networks was proposed for fall and posture position detection. This approach gave good results with a global accuracy of 93.44% and a global sensitivity of 93.24%. The detection of postures makes possible to follow the state of the person and in particular the behavior changes which are assumed to be the routine loss. Two strategies were deployed to monitor the routine. The first one examines the succession of activities in the day by computing an edit distance or a dynamic deformation of the day, the other one consists in classifying the day into routine and non-routine by combining supervised (k-means and k-modes), unsupervised (Random Forest) or a priori knowledge about the person's routine. These strategies were evaluated both on real data recorded in EHPAD in two frail people and on simulated data created to fill the lack of real data. They have shown the possibility to detect different behavioral change scenarios (abrupt, progressive, recurrent) and prove that depth sensors can be used in EHPAD or in the home of an elderly person
Allione, Julien. "Construction et validation d'un protocole visant à améliorer la détection du mensonge : une démarche de psychologie expérimentale appliquée." Toulouse 2, 2008. http://www.theses.fr/2008TOU20102.
Full textIn the last decades, research assessing the veracity of suspects statements, witnesses and alleged victims has become of great importance in eyewitness testimony scientifique literature. Indeed, statements are given a crucial role in eyewitness evidence. However, detecting a wrong statement is a difficult task, Several researches highlighted the inaccuracy for detecting false statement. Experimental studies tried to propose relevant cues to detect liar behaviour. Studies looking into verbal cues are a lot more promising than thoes interested in non-verbal cues. Two approaches analyzing principally written transcriptions of oral statements are currently subject to a lot of attention. One is the Criteria-Based Content Analysis (Steller and Köhnken, 1989) and the other is the Reality Monitoring (Johnson and Raye, 1989). Unfortunately, thoses approaches present few weakness. The first step of our research program consisted in extracting the most relevant verbal cues. In the same time, we tried to increase the criteria efficiency, using particular interviewing procedures. Then, we have submited the extracted cues to law offenders, in order to know if on the basis of transcription, those cues could allow better scoring in detecting sincere or untruthful testimonies. Finally, we tested if our approach would still be effective without having to retranscribe statements. The results show an improvment in detecting sincere to untruthfull testimonies
Nazir, Souha. "Evaluation d’un système de détection surfacique ‘Kinect V2’ dans différentes applications médicales." Thesis, Brest, 2018. http://www.theses.fr/2018BRES0101.
Full textIn recent years, one of the major technological innovations has been the introduction of depth cameras that can be used in a wide range of applications, including robotics, computer vision, automation, etc. These devices have opened up new opportunities for scientific research applied to the medical field. In this thesis, we will evaluate the potential use of the "Kinect V2" depth camera in order to respond to current clinical issues in radiotherapy and resuscitation in intensive care unit.Given that radiotherapy treatment is administered over several sessions, one of the key task is to daily reposition the patient in the same way as during the planning session.The precision of such repositioning is impacted by the respiratory motion. On the other hand, the movements of the machine as well as the possible movements of the patient can lead to machine / machine or machine /patient collisions. We propose a surface detection system for the management of inter and intra-fraction motion in external radiotherapy. This system is based on a rigid surface registration algorithm to estimate the treatment position and a real-time collision detection system to ensure patient safety during the treatment.Obtained results are encouraging and show a good agreement with available clinical systems.Concerning medical resuscitation, there is a need for new non-invasive and non-contact devices in order to optimize patient care. Non-invasive monitoring of spontaneous breathing for unstable patients is crucial in the intensive care unit. In this context, we propose a non-contact measurement system capable of calculating the parameters of patient's ventilation by observing thoracic morphological movements. The developed method gives a clinically acceptable precision. Such system is the first to solve previously described issue
Al, chanti Dawood. "Analyse Automatique des Macro et Micro Expressions Faciales : Détection et Reconnaissance par Machine Learning." Thesis, Université Grenoble Alpes (ComUE), 2019. http://www.theses.fr/2019GREAT058.
Full textFacial expression analysis is an important problem in many biometric tasks, such as face recognition, face animation, affective computing and human computer interface. In this thesis, we aim at analyzing facial expressions of a face using images and video sequences. We divided the problem into three leading parts.First, we study Macro Facial Expressions for Emotion Recognition and we propose three different levels of feature representations. Low-level feature through a Bag of Visual Word model, mid-level feature through Sparse Representation and hierarchical features through a Deep Learning based method. The objective of doing this is to find the most effective and efficient representation that contains distinctive information of expressions and that overcomes various challenges coming from: 1) intrinsic factors such as appearance and expressiveness variability and 2) extrinsic factors such as illumination, pose, scale and imaging parameters, e.g., resolution, focus, imaging, noise. Then, we incorporate the time dimension to extract spatio-temporal features with the objective to describe subtle feature deformations to discriminate ambiguous classes.Second, we direct our research toward transfer learning, where we aim at Adapting Facial Expression Category Models to New Domains and Tasks. Thus we study domain adaptation and zero shot learning for developing a method that solves the two tasks jointly. Our method is suitable for unlabelled target datasets coming from different data distributions than the source domain and for unlabelled target datasets with different label distributions but sharing the same context as the source domain. Therefore, to permit knowledge transfer between domains and tasks, we use Euclidean learning and Convolutional Neural Networks to design a mapping function that map the visual information coming from facial expressions into a semantic space coming from a Natural Language model that encodes the visual attribute description or use the label information. The consistency between the two subspaces is maximized by aligning them using the visual feature distribution.Third, we study Micro Facial Expression Detection. We propose an algorithm to spot micro-expression segments including the onset and offset frames and to spatially pinpoint in each image space the regions involved in the micro-facial muscle movements. The problem is formulated into Anomaly Detection due to the fact that micro-expressions occur infrequently and thus leading to few data generation compared to natural facial behaviours. In this manner, first, we propose a deep Recurrent Convolutional Auto-Encoder to capture spatial and motion feature changes of natural facial behaviours. Then, a statistical based model for estimating the probability density function of normal facial behaviours while associating a discriminating score to spot micro-expressions is learned based on a Gaussian Mixture Model. Finally, an adaptive thresholding technique for identifying micro expressions from natural facial behaviour is proposed.Our algorithms are tested over deliberate and spontaneous facial expression benchmarks
Brazey, Denis. "Reconnaissance de formes et suivi de mouvements en 4D temps-réel : Restauration de cartes de profondeur." Thesis, Rouen, INSA, 2014. http://www.theses.fr/2014ISAM0019.
Full textIn this dissertation, we are interested in several issues related to 3D data processing. The first one concerns people detection and tracking in depth map sequences. We propose an improvement of an existing method based on a segmentation stage followed by a tracking module. The second issue is head detection and modelling in 3D point clouds. In order to do this, we adopt a probabilistic approach based on a new spherical mixture model. The last considered application deals with the restoration of deteriorated depth maps. To solve this problem, we propose to use a surface approximation method based on interpolation Dm-splines with scale transforms to approximate and restore the image. Presented results illustrate the efficiency of the developed algorithms
Artaud, Chloé. "Détection des fraudes : de l’image à la sémantique du contenu : application à la vérification des informations extraites d’un corpus de tickets de caisse." Thesis, La Rochelle, 2019. http://www.theses.fr/2019LAROS002/document.
Full textCompanies, administrations, and sometimes individuals, have to face many frauds on documents they receive from outside or process internally. Invoices, expense reports, receipts...any document used as proof can be falsified in order to earn more money or not to lose it. In France, losses due to fraud are estimated at several billion euros per year. Since the flow of documents exchanged, whether digital or paper, is very important, it would be extremely costly and time-consuming to have them all checked by fraud detection experts. That’s why we propose in our thesis a system for automatic detection of false documents. While most of the work in automatic document detection focuses on graphic clues, we seek to verify the textual information in the document in order to detect inconsistencies or implausibilities.To do this, we first compiled a corpus of documents that we digitized. After correcting the characters recognition outputs and falsifying part of the documents, we extracted the information and modelled them in an ontology, in order to keep the semantic links between them. The information thus extracted, and increased by its possible disambiguation, can be verified against each other within the document and through the knowledge base established. The semantic links of ontology also make it possible to search for information in other sources of knowledge, particularly on the Internet
Dubois, Amandine. "Mesure de la fragilité et détection de chutes pour le maintien à domicile des personnes âgées." Phd thesis, Université de Lorraine, 2014. http://tel.archives-ouvertes.fr/tel-01070972.
Full textAuvinet, Edouard. "Analyse d’information tridimensionnelle issue de systèmes multi-caméras pour la détection de la chute et l’analyse de la marche." Thèse, Rennes 2, 2012. http://hdl.handle.net/1866/9770.
Full textThis thesis is concerned with defining new clinical investigation method to assess the impact of ageing on motricity. In particular, this thesis focuses on two main possible disturbance during ageing : the fall and walk impairment. This two motricity disturbances still remain unclear and their clinical analysis presents real scientist and technological challenges. In this thesis, we propose novel measuring methods usable in everyday life or in the walking clinic, with a minimum of technical constraints. In the first part, we address the problem of fall detection at home, which was widely discussed in previous years. In particular, we propose an approach to exploit the subject’s volume, reconstructed from multiple calibrated cameras. These methods are generally very sensitive to occlusions that inevitably occur in the home and we therefore propose an original approach much more robust to these occultations. The efficiency and real-time operation has been validated on more than two dozen videos of falls and lures, with results approaching 100 % sensitivity and specificity with at least four or more cameras. In the second part, we go a little further in the exploitation of reconstructed volumes of a person at a particular motor task : the treadmill, in a clinical diagnostic. In this section we analyze more specifically the quality of walking. For this we develop the concept of using depth camera for the quantification of the spatial and temporal asymmetry of lower limb movement during walking. After detecting each step in time, this method makes a comparison of surfaces of each leg with its corresponding symmetric leg in the opposite step. The validation performed on a cohort of 20 subjects showed the viability of the approach.
Réalisé en cotutelle avec le laboratoire M2S de Rennes 2
Le, François Josette. "Une lecture théologique de la conversion chez Thomas Merton à travers la triple clé biblique de Mc 8, 34; Jn 3, 7 et Ga 2, 20a la théologie de la conversion chez Thomas Merton." Thèse, Université de Sherbrooke, 2009. http://savoirs.usherbrooke.ca/handle/11143/5224.
Full textHofmanis, Janis. "Contribution au modèle direct cérébral par stimulation électrique de profondeur et mesures SEEG : application à l'épilepsie." Thesis, Université de Lorraine, 2013. http://www.theses.fr/2013LORR0209/document.
Full textThe study of epilepsy requires the identification of cerebral structures which are involved in generation of seizures and connexion processes. Several methods of clinical investigation contributed to these studies : imaging (PET, MRI), electrophysiology (EEG, SEEG, MEG). The EEG provides a temporal resolution enough to analyze these processes. However, the localization of deep sources and their dynamical properties are difficult to understand. SEEG is a modality of intracerebral electrophysiological and anatomical high temporal resolution reserved for some difficult cases of pre-surgical diagnosis : drug-resistant epilepsy. The definition of the epileptogenic zone, as proposed by Talairach and Bancaud is an electro-clinical definition based on the results of intracerebral SEEG recordings. It takes into account not only the anatomical localization of partial epileptic discharge, but also the dynamic evolution of this discharge (active neural networks at the time of seizure) and clinical symptoms. Recently, a novel diagnostic technique allows an accurate localization of the epileptogenic zone using Depth Brain Stimulation (DBS). This exogenous source can activate the epileptic networks and generate an electrophysiological reaction. Therefore, coupling DBS with SEEG measurements is very advantageous : firstly, to contribute to the modeling and understanding of the (epileptic) brain and to help the diagnosis, secondly, to access the estimation of head model as an electrical conductor (conductive properties of tissues). In addition, supplementary information about head model improves the solution to the inverse problem (source localization methods) used in many applications in EEG and SEEG. The inverse solution requires repeated computation of the forward problem, i.e. the simulation of EEG and SEEG fields for a given dipolar source in the brain using a volume-conduction model of the head. As for DBS, the location of source is well defined. Therefore, in this thesis, we search for the best head model for the forward problem from real synchronous measurements of EEG and SEEG with DBS in several patients. So, the work of the thesis breaks up into different parts for which we need to accomplish the following tasks : Creation of database 3000 DBS measurements for 42 patients ; Extraction of DBS signal from SEEG and EEG measurements using multidimensional analysis : 5 methods have been developed or adapted and validate first in a simulation study and, secondly, in a real SEEG application ; Localization of SEEG electrodes in MR and CT images, including segmentation of brain matter ; SEEG forward modeling using infinite medium, spherical and realistic models based on MRI and CT of the patient ; Comparison between different head models and validation with real in vivo DBS measurements ; Validation of realistic 5-compartment FEM head models by incorporating the conductivities of cerebrospinal fluid (CSF), gray and white matters
Poulain, d'Andecy Vincent. "Système à connaissance incrémentale pour la compréhension de document et la détection de fraude." Thesis, La Rochelle, 2021. http://www.theses.fr/2021LAROS025.
Full textThe Document Understanding is the Artificial Intelligence ability for machines to Read documents. In a global vision, it aims the understanding of the document function, the document class, and in a more local vision, it aims the understanding of some specific details like entities. The scientific challenge is to recognize more than 90% of the data. While the industrial challenge requires this performance with the least human effort to train the machine. This thesis defends that Incremental Learning methods can cope with both challenges. The proposals enable an efficient iterative training with very few document samples. For the classification task, we demonstrate (1) the continue learning of textual descriptors, (2) the benefit of the discourse sequence, (3) the benefit of integrating a Souvenir of few samples in the knowledge model. For the data extraction task, we demonstrate an iterative structural model, based on a star-graph representation, which is enhanced by the embedding of few a priori knowledges. Aware about economic and societal impacts because the document fraud, this thesis deals with this issue too. Our modest contribution is only to study the different fraud categories to open further research. This research work has been done in a non-classic framework, in conjunction of industrial activities for Yooz and collaborative research projects like the FEDER Securdoc project supported by la région Nouvelle Aquitaine, and the Labcom IDEAS supported by the ANR
Ok, David. "Mise en Correspondance Robuste et Détection d'Éléments Visuels Appliquées à l'Analyse de Façades." Phd thesis, Université Paris-Est, 2013. http://tel.archives-ouvertes.fr/tel-00844049.
Full textAuvinet, Edouard. "Analyse d'information tridimensionnelle issue de systèmes multi-caméras pour la détection de la chute et l'analyse de la marche." Phd thesis, Université Rennes 2, 2012. http://tel.archives-ouvertes.fr/tel-00946188.
Full textWang, Qiong. "Salient object detection and segmentation in videos." Thesis, Rennes, INSA, 2019. http://www.theses.fr/2019ISAR0003/document.
Full textThis thesis focuses on the problem of video salient object detection and video object instance segmentation which aim to detect the most attracting objects or assign consistent object IDs to each pixel in a video sequence. One approach, one overview and one extended model are proposed for video salient object detection, and one approach is proposed for video object instance segmentation. For video salient object detection, we propose: (1) one traditional approach to detect the whole salient object via the adjunction of virtual borders. A guided filter is applied on the temporal output to integrate the spatial edge information for a better detection of the salient object edges. A global spatio-temporal saliency map is obtained by combining the spatial saliency map and the temporal saliency map together according to the entropy. (2) An overview of recent developments for deep-learning based methods is provided. It includes the classifications of the state-of-the-art methods and their frameworks, and the experimental comparison of the performances of the state-of-the-art methods. (3) One extended model further improves the performance of the proposed traditional approach by integrating a deep-learning based image salient object detection method For video object instance segmentation, we propose a deep-learning approach in which the warping confidence computation firstly judges the confidence of the mask warped map, then a semantic selection is introduced to optimize the warped map, where the object is re-identified using the semantics labels of the target object. The proposed approaches have been assessed on the published large-scale and challenging datasets. The experimental results show that the proposed approaches outperform the state-of-the-art methods
Hoarau, Gwenaël. "Étude de la limite de détection et des fausses alarmes émises par les moniteurs de mesure de la contamination radioactive atmosphérique dans les chantiers de démantèlement." Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASP055.
Full textThe Continuous Air monitors are used in the nuclear facilities for ensuring the radiation protection of workers who are likely to be confronted to radioactive aerosols. The CAM behavior are evaluated under IEC normative conditions. In atmospheric conditions as measured in decommissioning nuclear sites, outside of IEC, wrongs CAM behavior are observed. Which correspond to the false positives triggering. The aim of this thesis work is to enhance the knowledge about the CAM behavior when it is faced to outside IEC normative conditions. A specific experimental chamber has been designed, inside which the reproduction of the dismantling nuclear site atmosphere is achievable. Thus, this experimental chamber made it possible, in the laboratory, the study of the behavior of a CAM type ABPM203M. With the results we have, on the one hand, highlighted the conditions which lead to unexpected behavior of the CAM and, on the other hand, demonstrated the reasons why a false positive alarm is often emitted under these conditions, outside IEC. At the synthesis of the results of the study, a new measurement process was proposed, which makes it possible, on the one hand, to consider both the characteristics of non-radioactive aerosols and those of radioactive aerosols and on the other hand, to improve the measurement achieved out by the CAM to always ensure the radiation protection of workers
Manceau, Jérôme. "Clonage réaliste de visage." Thesis, CentraleSupélec, 2016. http://www.theses.fr/2016SUPL0004/document.
Full text3D face clones can be used in many areas such as Human-Computer Interaction and as pretreatment in applications such as emotion analysis. However, such clones should have well-modeled facial shape while keeping the specificities of individuals and they should be semantic. A clone is semantic when we know the position of the different parts of the face (eyes, nose...). In our technique, we use a RGB-D sensor to get the specificities of individuals and 3D Morphable Face Model to mark facial shape. For the reconstruction of the shape, we reverse the process classically used. Indeed, we first perform fitting and then data fusion. For each depth frame, we keep the suitable parts of data called patches. Depending on the location, we merge either sensor data or 3D Morphable Face Model data. For the reconstruction of the texture, we use shape and texture patches to preserve the person's characteristics. They are detected using the depth frames of a RGB-D sensor. The tests we perform show the robustness and the accuracy of our method
Leh, Barbara. "Caractérisation par autofluorescence de tissus cérébraux tumoraux : mesures sur fantômes et modèle animal." Phd thesis, Université Paris Sud - Paris XI, 2011. http://tel.archives-ouvertes.fr/tel-00647327.
Full textNguyen, Hoai phuong. "Certification de l'intégrité d'images numériques et de l'authenticité." Thesis, Reims, 2019. http://www.theses.fr/2019REIMS007/document.
Full textNowadays, with the advent of the Internet, the falsification of digital media such as digital images and video is a security issue that cannot be ignored. It is of vital importance to certify the conformity and the integrity of these media. This project, which is in the domain of digital forensics, is proposed to answer this problematic
Bourien, Jérôme. "Analyse de distributions spatio-temporelles de transitoires dans des signaux vectoriels. Application à la détection-classification d'activités paroxystiques intercritiques dans des observations EEG." Phd thesis, Université Rennes 1, 2003. http://tel.archives-ouvertes.fr/tel-00007178.
Full text1. Détection des AE monovoie. La méthode de détection, qui repose sur une approche heuristique, utilise un banc de filtres en ondelettes pour réhausser la composante pointue des AE (généralement appelée "spike" dans la littérature). La valeur moyenne des statistiques obtenues en sortie de chaque filtre est ensuite analysée avec un algorithme de Page-Hinkley dans le but de détecter des changements abrupts correspondant aux spikes.
2. Fusion des AE. Cette procédure recherche des co-occurrences entre AE monovoie à l'aide d'une fenêtre glissante puis forme des AE multivoies.
3. Extraction des sous-ensembles de voies fréquement et significativement activées lors des AE multivoies (appelés "ensembles d'activation").
4. Evaluation de l'éxistence d'un ordre d'activation temporel reproductible (éventuellement partiel) au sein de chaque ensemble d'activation.
Les méthodes proposées dans chacune des étapes ont tout d'abord été évaluées à l'aide de signaux simulés (étape 1) ou à l'aide de models Markoviens (étapes 2-4). Les résultats montrent que la méthode complète est robuste aux effets des fausses-alarmes. Cette méthode a ensuite été appliquée à des signaux enregistrés chez 8 patients (chacun contenant plusieurs centaines d'AE). Les résultats indiquent une grande reproductibilité des distributions spatio-temporelles des AE et ont permis l'identification de réseaux anatomo-fonctionnels spécifiques.
Billiot, Bastien. "Conception d'un dispositif d'acquisition d'images agronomiques 3D en extérieur et développement des traitements associés pour la détection et la reconnaissance de plantes et de maladies." Phd thesis, Université de Bourgogne, 2013. http://tel.archives-ouvertes.fr/tel-00983327.
Full textLaborde, Antoine. "Detection of minor compounds in food powder using near infrared hyperspectral imaging." Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASB017.
Full textNear-infrared (NIR) hyperspectral imaging provides a spectral map for organic samples. Minor compounds in food powder can be looked for by analyzing the pixel spectra. However, the NIR spectral analysis is limited to a given depth. Besides, particles smaller than the pixel size induce a mixed spectral signature in the pixels. These two issues are an obstacle to the analysis of minor compounds in food powders.We propose a method to determine the detection depth of a composite target under a layer of powder such as wheat flour. It is based on the Partial Least Squares regression and provides an understanding of how the NIR signal is attenuated when the layer of powder despite the penetration depth issues.Two spectral unmixing strategies are proposed to detect pixel with minor compound NIR signatures. The lack of reference values to validate the model and the ambiguity of the spectral signature to unmix are two major difficulties. The first method models the spectral variability using Principal Component Analysis to design a performant detection algorithm. Then, for a more complex situation, the Multivariate Curve Resolution Alternating Least-Squares algorithm is used to unmix each pixel
Derome, Maxime. "Vision stéréoscopique temps-réel pour la navigation autonome d'un robot en environnement dynamique." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLS156/document.
Full textThis thesis aims at designing an embedded stereoscopic perception system that enables autonomous robot navigation in dynamic environments (i.e. including mobile objects). To do so, we need to satisfy several constraints: 1) We want to be able to navigate in unknown environment and with any type of mobile objects, thus we adopt a geometric approach. 2) We want to ensure the best possible coverage of the field of view, so we employ dense methods that process every pixel in the image. 3) The algorithms must be compliant with an embedded platform, therefore we must carefully design the algorithms so they are fast enough to keep a certain level of reactivity. The approach presented in this thesis manuscript and the contributions are summarized below. First, we study several stereo matching algorithms that estimate a disparity map from which we can deduce a depth map, by triangulation. This comparative study highlights one algorithm that is not in the KITTI benchmarks, but that gives a great accuracy/processing time tradeoff. We also propose a filtering method to post-process the disparity maps. By coding these algorithm in CUDA to benefit from hardware acceleration on Graphics Processing Unit, we show that they can perform very fast (19ms on KITTI images, with a GPU GeForce GTX Titan).Second, we want to detect mobile objects and estimate their motion. To do so we compute the stereo rig motion using visual odometry, in order to isolate the part induced by moving objects in the 2D or 3D apparent motion (estimated by optical flow or scene flow algorithms). Considering that the only optical flow algorithm able to perform in real-time is FOLKI, we propose several modifications of it to slightly improve its performances at the cost of a slower processing time. Regarding the scene flow estimation, existing algorithms cannot reach the desired computation speed, so we propose a new approach by decoupling structure and motion for a fast scene flow estimation. Three algorithms are proposed to use this structure-motion decomposition, and one of them, particularly efficient, enables very fast scene flow computing with a relatively good accuracy. To our knowledge it is the only published scene flow algorithm able to perform at framerate on KITTI dataset (10 Hz).Third, to detect moving objects and segment them in the image, we show several statistical models and residual quantities on which we can base the detection by thresholding a chi2 criterion. We propose a rigorous statistical modeling that takes into account all the uncertainties occurring during the estimation, in particular during the visual odometry, which had not been done to our knowledge, in the context of moving object detection. We also propose a new residual quantity for the detection, using an image prediction approach to facilitate uncertainty propagation and the chi2 criterion modeling. The benefit brought by the proposed residual quantity and error model is demonstrated by evaluating detection algorithms on a samples of annotated KITTI data. Finally, we implement our algorithms on ROS to run the perception system on en embedded platform, and we code some algorithms in CUDA to accelerate the computing using GPU. We describe the perception and the navigation system that we use for the experimental validation. We show in our experiments that the proposed stereovision perception system is suitable for embedded robotic applications
Mozafari, Marzieh. "Hate speech and offensive language detection using transfer learning approaches." Electronic Thesis or Diss., Institut polytechnique de Paris, 2021. http://www.theses.fr/2021IPPAS007.
Full textThe great promise of social media platforms (e.g., Twitter and Facebook) is to provide a safe place for users to communicate their opinions and share information. However, concerns are growing that they enable abusive behaviors, e.g., threatening or harassing other users, cyberbullying, hate speech, racial and sexual discrimination, as well. In this thesis, we focus on hate speech as one of the most concerning phenomenon in online social media.Given the high progression of online hate speech and its severe negative effects, institutions, social media platforms, and researchers have been trying to react as quickly as possible. The recent advancements in Natural Language Processing (NLP) and Machine Learning (ML) algorithms can be adapted to develop automatic methods for hate speech detection in this area.The aim of this thesis is to investigate the problem of hate speech and offensive language detection in social media, where we define hate speech as any communication criticizing a person or a group based on some characteristics, e.g., gender, sexual orientation, nationality, religion, race. We propose different approaches in which we adapt advanced Transfer Learning (TL) models and NLP techniques to detect hate speech and offensive content automatically, in a monolingual and multilingual fashion.In the first contribution, we only focus on English language. Firstly, we analyze user-generated textual content to gain a brief insight into the type of content by introducing a new framework being able to categorize contents in terms of topical similarity based on different features. Furthermore, using the Perspective API from Google, we measure and analyze the toxicity of the content. Secondly, we propose a TL approach for identification of hate speech by employing a combination of the unsupervised pre-trained model BERT (Bidirectional Encoder Representations from Transformers) and new supervised fine-tuning strategies. Finally, we investigate the effect of unintended bias in our pre-trained BERT based model and propose a new generalization mechanism in training data by reweighting samples and then changing the fine-tuning strategies in terms of the loss function to mitigate the racial bias propagated through the model. To evaluate the proposed models, we use two publicly available datasets from Twitter.In the second contribution, we consider a multilingual setting where we focus on low-resource languages in which there is no or few labeled data available. First, we present the first corpus of Persian offensive language consisting of 6k micro blog posts from Twitter to deal with offensive language detection in Persian as a low-resource language in this domain. After annotating the corpus, we perform extensive experiments to investigate the performance of transformer-based monolingual and multilingual pre-trained language models (e.g., ParsBERT, mBERT, XLM-R) in the downstream task. Furthermore, we propose an ensemble model to boost the performance of our model. Then, we expand our study into a cross-lingual few-shot learning problem, where we have a few labeled data in target language, and adapt a meta-learning based approach to address identification of hate speech and offensive language in low-resource languages
Alla, Jules-Ryane S. "Détection de chute à l'aide d'une caméra de profondeur." Thèse, 2013. http://hdl.handle.net/1866/9992.
Full textElderly falls are a major public health problem. Studies show that about 30% of people aged 65 and older fall each year in Canada, with negative consequences on individuals, their families and our society. Faced with such a situation a video surveillance system is an effective solution to ensure the safety of these people. To this day many systems support services to the elderly. These devices allow the elderly to live at home while ensuring their safety by wearing a sensor. However the sensor must be worn at all times by the subject which is uncomfortable and restrictive. This is why research has recently been interested in the use of cameras instead of wearable sensors. The goal of this project is to demonstrate that the use of a video surveillance system can help to reduce this problem. In this thesis we present an approach for automatic detection of falls based on a method for tracking 3D subject using a depth camera (Kinect from Microsoft) positioned vertically to the ground. This monitoring is done using the silhouette extracted in real time with a robust approach for extracting 3D depth based on the depth variation of the pixels in the scene. This method is based on an initial capture the scene without any body. Once extracted, 10% of the silhouette corresponding to the uppermost region (nearest to the Kinect) will be analyzed in real time depending on the speed and the position of its center of gravity . These criteria will be analysed to detect the fall, then a signal (email or SMS) will be transmitted to an individual or to the authority in charge of the elderly. This method was validated using several videos of a stunt simulating falls. The camera position and depth information reduce so considerably the risk of false alarms. Positioned vertically above the ground, the camera makes it possible to analyze the scene especially for tracking the silhouette without major occlusion, which in some cases lead to false alarms. In addition, the various criteria for fall detection, are reliable characteristics for distinguishing the fall of a person, from squatting or sitting. Nevertheless, the angle of the camera remains a problem because it is not large enough to cover a large surface. A solution to this dilemma would be to fix a lens on the objective of the Kinect for the enlargement of the field of view and monitored area.
Ndayikengurukiye, Didier. "Estimation de cartes d'énergie de hautes fréquences ou d'irrégularité de périodicité de la marche humaine par caméra de profondeur pour la détection de pathologies." Thèse, 2016. http://hdl.handle.net/1866/16178.
Full textThis work presents two new and simple human gait analysis systems based on a depth camera (Microsoft Kinect) placed in front of a subject walking on a conventional treadmill, capable of detecting a healthy gait from an impaired one. The first system presented relies on the fact that a normal walk typically exhibits a smooth motion (depth) signal, at each pixel with less high-frequency spectral energy content than an abnormal walk. This permits to estimate a map for that subject, showing the location and the amplitude of the high-frequency spectral energy (HFSE). The second system analyses the patient's body parts that have an irregular movement pattern, in terms of periodicity, during walking. Herein we assume that the gait of a healthy subject exhibits anywhere in the human body, during the walking cycles, a depth signal with a periodic pattern without noise. From each subject’s video sequence, we estimate a saliency color map showing the areas of strong gait irregularities also called aperiodic noise energy. Either the HFSE or aperiodic noise energy shown in the map can be used as a good indicator of possible pathology in an early, fast and reliable diagnostic tool or to provide information about the presence and extent of disease or (orthopedic, muscular or neurological) patient's problems. Even if the maps obtained are informative and highly discriminant for a direct visual classification, even for a non-specialist, the proposed systems allow us to automatically detect maps representing healthy individuals and those representing individuals with locomotor problems.
Gros, D'Aillon Eric. "Etude des performances spectrometriques des detecteurs gamma CdTe CdZnTe monolithiques." Phd thesis, 2005. http://tel.archives-ouvertes.fr/tel-00011119.
Full textCe travail a consisté à étudier expérimentalement et par simulation, les corrélations entre le pas des anodes, les propriétés physiques des matériaux (résistivité et propriétés de transport des électrons) et les performances spectrométriques des détecteurs. Nous avons comparé plusieurs méthodes de mesure de la profondeur d'interaction des photons et avons obtenu une résolution en énergie à 122 keV comprise entre 1.7 % et 7 %, selon le matériau, pour 5 mm d'épaisseur. Le partage de charges entre les anodes des détecteurs a été étudié et un traitement des informations mesurées est proposé.