Tesis sobre el tema "Classification des réseaux de neurones"
Crea una cita precisa en los estilos APA, MLA, Chicago, Harvard y otros
Consulte los 50 mejores tesis para su investigación sobre el tema "Classification des réseaux de neurones".
Junto a cada fuente en la lista de referencias hay un botón "Agregar a la bibliografía". Pulsa este botón, y generaremos automáticamente la referencia bibliográfica para la obra elegida en el estilo de cita que necesites: APA, MLA, Harvard, Vancouver, Chicago, etc.
También puede descargar el texto completo de la publicación académica en formato pdf y leer en línea su resumen siempre que esté disponible en los metadatos.
Explore tesis sobre una amplia variedad de disciplinas y organice su bibliografía correctamente.
Biela, Philippe. "Classification automatique d'observations multidimensionnelles par réseaux de neurones compétitifs". Lille 1, 1999. https://pepite-depot.univ-lille.fr/LIBRE/Th_Num/1999/50376-1999-469.pdf.
Texto completoChakik, Fadi El. "Maximum d'entropie et réseaux de neurones pour la classification". Grenoble INPG, 1998. http://www.theses.fr/1998INPG0091.
Texto completoAyache, Mohammad. "Application des réseaux de neurones à la classification automatisée des grades placentaires". Tours, 2007. http://www.theses.fr/2007TOUR3315.
Texto completoThe placenta is a temporary organ joins the mother and the fœtus, which transfers oxygen from the mother to the foetus, allows the evacuation of the carbon dioxide and the products of foetus metabolism. The goal of our work is to study the transfer function of placental development using ultrasound images. A new approach is developed during this work to classify the placental development by image processing techniques based on supervised neural network. The realized model by the wavelet transform based on MLP neural network, represents an effective tool answering our criteria and adapted to our applications concerning the study of placental maturation. The realized model application in the event of placental image processing opens interesting doors in terms of placental grades classification in order to identify the stages of maturation, authorizing the definition of a normal maturation and an abnormal maturation
Zaki, Sabit Fawzi Philippe. "Classification par réseaux de neurones dans le cadre de la scattérométrie ellipsométrique". Thesis, Lyon, 2016. http://www.theses.fr/2016LYSES070/document.
Texto completoThe miniaturization of components in the micro-electronics industry involves the need of fast reliable technique of characterization with lower cost. Optical methods such as scatterometry are today promising alternative to this technological need. However, scatterometric method requires a certain number of hypothesis to ensure the resolution of an inverse problem, in particular the knowledge of the geometrical shape of the structure under test. The assumed model of the structure determines the quality of the characterization. In this thesis, we propose the use of neural networks as decision-making tools upstream of any characterization method. We validated the use of neural networks in the context of recognition of the geometrical shapes of the sample under testing by the use of optical signature in any scatterometric characterization process. First, the case of lithographic defect due to the presence of a resist residual layer at the bottom of the grooves is studied. Then, we carry out an analysis of model defect in the inverse problem resolution. Finally, we report results in the context of selection of geometric models by neural networks upstream of a classical scatterometric characterization process. This thesis has demonstrated that neural networks can well answer the problem of classification in ellipsometric scatterometry and their use can improve this optical characterization technique
Gatet, Laurent. "Intégration de Réseaux de Neurones pour la Télémétrie Laser". Phd thesis, Toulouse, INPT, 2007. http://oatao.univ-toulouse.fr/7595/1/gatet.pdf.
Texto completoDelsert, Stéphane. "Classification interactive non supervisée de données multidimensionnelles par réseaux de neurones à apprentissage cométitif". Lille 1, 1996. https://pepite-depot.univ-lille.fr/LIBRE/Th_Num/1996/50376-1996-214.pdf.
Texto completoBouaziz, Mohamed. "Réseaux de neurones récurrents pour la classification de séquences dans des flux audiovisuels parallèles". Thesis, Avignon, 2017. http://www.theses.fr/2017AVIG0224/document.
Texto completoIn the same way as TV channels, data streams are represented as a sequence of successive events that can exhibit chronological relations (e.g. a series of programs, scenes, etc.). For a targeted channel, broadcast programming follows the rules defined by the channel itself, but can also be affected by the programming of competing ones. In such conditions, event sequences of parallel streams could provide additional knowledge about the events of a particular stream. In the sphere of machine learning, various methods that are suited for processing sequential data have been proposed. Long Short-Term Memory (LSTM) Recurrent Neural Networks have proven its worth in many applications dealing with this type of data. Nevertheless, these approaches are designed to handle only a single input sequence at a time. The main contribution of this thesis is about developing approaches that jointly process sequential data derived from multiple parallel streams. The application task of our work, carried out in collaboration with the computer science laboratory of Avignon (LIA) and the EDD company, seeks to predict the genre of a telecast. This prediction can be based on the histories of previous telecast genres in the same channel but also on those belonging to other parallel channels. We propose a telecast genre taxonomy adapted to such automatic processes as well as a dataset containing the parallel history sequences of 4 French TV channels. Two original methods are proposed in this work in order to take into account parallel stream sequences. The first one, namely the Parallel LSTM (PLSTM) architecture, is an extension of the LSTM model. PLSTM simultaneously processes each sequence in a separate recurrent layer and sums the outputs of each of these layers to produce the final output. The second approach, called MSE-SVM, takes advantage of both LSTM and Support Vector Machines (SVM) methods. Firstly, latent feature vectors are independently generated for each input stream, using the output event of the main one. These new representations are then merged and fed to an SVM algorithm. The PLSTM and MSE-SVM approaches proved their ability to integrate parallel sequences by outperforming, respectively, the LSTM and SVM models that only take into account the sequences of the main stream. The two proposed approaches take profit of the information contained in long sequences. However, they have difficulties to deal with short ones. Though MSE-SVM generally outperforms the PLSTM approach, the problem experienced with short sequences is more pronounced for MSE-SVM. Finally, we propose to extend this approach by feeding additional information related to each event in the input sequences (e.g. the weekday of a telecast). This extension, named AMSE-SVM, has a remarkably better behavior with short sequences without affecting the performance when processing long ones
Carpentier, Mathieu. "Classification fine par réseau de neurones à convolution". Master's thesis, Université Laval, 2019. http://hdl.handle.net/20.500.11794/35835.
Texto completoArtificial intelligence is a relatively recent research domain. With it, many breakthroughs were made on a number of problems that were considered very hard. Fine-grained classification is one of those problems. However, a relatively small amount of research has been done on this task even though itcould represent progress on a scientific, commercial and industrial level. In this work, we talk about applying fine-grained classification on concrete problems such as tree bark classification and mould classification in culture. We start by presenting fundamental deep learning concepts at the root of our solution. Then, we present multiple experiments made in order to try to solve the tree bark classification problem and we detail the novel dataset BarkNet 1.0 that we made for this project. With it, we were able to develop a method that obtains an accuracy of 93.88% on singlecrop in a single image, and an accuracy of 97.81% using a majority voting approach on all the images of a tree. We conclude by demonstrating the feasibility of applying our method on new problems by showing two concrete applications on which we tried our approach, industrial tree classification and mould classification.
Mercadier, Yves. "Classification automatique de textes par réseaux de neurones profonds : application au domaine de la santé". Thesis, Montpellier, 2020. http://www.theses.fr/2020MONTS068.
Texto completoThis Ph.D focuses on the analysis of textual data in the health domain and in particular on the supervised multi-class classification of data from biomedical literature and social media.One of the major difficulties when exploring such data by supervised learning methods is to have a sufficient number of data sets for models training. Indeed, it is generally necessary to label manually the data before performing the learning step. The large size of the data sets makes this labellisation task very expensive, which should be reduced with semi-automatic systems.In this context, active learning, in which the Oracle intervenes to choose the best examples to label, is promising. The intuition is as follows: by choosing the smartly the examples and not randomly, the models should improve with less effort for the oracle and therefore at lower cost (i.e. with less annotated examples). In this PhD, we will evaluate different active learning approaches combined with recent deep learning models.In addition, when small annotated data set is available, one possibility of improvement is to artificially increase the data quantity during the training phase, by automatically creating new data from existing data. More precisely, we inject knowledge by taking into account the invariant properties of the data with respect to certain transformations. The augmented data can thus cover an unexplored input space, avoid overfitting and improve the generalization of the model. In this Ph.D, we will propose and evaluate a new approach for textual data augmentation.These two contributions will be evaluated on different textual datasets in the medical domain
Personnaz, Léon. "Etude des réseaux de neurones formels : conception, propriétés et applications". Paris 6, 1986. http://www.theses.fr/1986PA066569.
Texto completoThiaw, Lamine. "Identification de systèmes dynamiques non linéaires par réseaux de neurones et multimodèles". Phd thesis, Université Paris XII Val de Marne, 2008. http://tel.archives-ouvertes.fr/tel-00399469.
Texto completoGalerne, Pascal. "Détection et classification de cibles posées sur le fond marin par réseaux de neurones en imagerie sonar". Brest, 1998. http://www.theses.fr/1998BRES2022.
Texto completoBensekka, Chakib. "Approche topologique de la métrologie du mouvement pour des applications en réalité virtuelle". Thesis, Paris, ENSAM, 2018. http://www.theses.fr/2018ENAM0040/document.
Texto completoIn the medical field, a better knowledge of the motor function isimportant for us to determine therapies adapted to each motor lesion andtools of studies and screening for neurodegenerative diseases. In thedomain of virtual reality, motion recognition is an issue in theinteraction of the avatar or the user in immersion with theirenvironment.Several studies have been conducted with the aim of proposingapproaches to the classification of human movement. The main idea ofthese methods is to extract invariants from the recorded data in orderto order them into clusters. However, the study of human motion withmotion capture systems generates a big quantity of data with nonlinearrelations between them. The presented methods in the scientificliterature use these data either directly as input to classificationalgorithms or by applying a dimensional reduction method such asprincipal component analysis prior to classification. These methodsremain extremely sensitive to white noise during recording as well asmorphological differences between subjects.In our work, we will present a methodology of classification andrecognition of human movement which is based on the topologicalanalysis of kinematic data. Topological analysis will be performed viahomological persistence which is a large data analysis method thatallows them to be topologically signed. This method of topologicalanalysis will be combined with learning algorithms to increase theaccuracy of motion recognition by reducing the impact of morphologicaldifferences between subjects, as well as the impact of white noiseissued during the step of movement acquisition. Also, we will combinethe topological analysis method with a temporal neural networkalgorithm in order to build an approach that allows to predict thecontinuation of a movement from a part of a recording interval.The results showed the ability of the proposed approach to achievehigh accuracy at classification, as well as its robustness againstwhite noise and morphological differences between subjects. Theresults also showed the high cost in computing time of our approachwhich we tried to reduce by modifying its steps and by rewriting thecode so that it can be executed in parallel
Jouni, Hassan. "Cellules analogiques CMOS pour réseaux de neurones. Application à la classification des cellules cancéreuses dans le sein". Thesis, Université Côte d'Azur (ComUE), 2018. http://www.theses.fr/2018AZUR4247.
Texto completoThe artificial neural networks are particularly interesting for CMOS VLSI (Very Large Scale Integration Complementary Metal-Oxide Semiconductor) implementations because every parallel element (neuron or synapsis) is relatively simple, allowing the complete integration of big networks on a single chip). Multipliers, non-linear function and its derivative are essential key elements in the analog signal processing in particular for analog VLSI implementation of artificial neuronal networks. The main conditions of this kind of circuits are the following ones: a low surface of Silicon and a low electric consumption. To validate our approach, we chose as type of application, the classification of cancer cells (malignant or benign) of the breast. There are many types of neural networks: Feed-forward neural network with back propagation (MLP), Radial Basis Network (RBN), Recurrent Neural Network (RNN) and other. The neural network studied in this thesis is based on Multi-Layer Perceptron with back-propagation (MLP).The main objective is to find the best compromises and the optimizations to realize circuits in a mature STMicroelectronics HCMOS9A 130nm technology and supplied with ± 900mV to have the lowest cost. Having chosen the best algorithm (the simplest and most effective) as a simple VLSI implementation, we defined efficient analog architecture. Finally building blocks were designed and realized before the final integration on a low surface of silicon and low power consumption. To verify and validate the project of the VLSI chip before manufacturing, a methodology of check was proposed in this thesis. It also allowed us to define the specifications of the full chip, as well as that of the building blocks
Villa-Vialaneix, Nathalie. "Eléments d'apprentissage en statistique fonctionnelle : classification et régression fonctionnelles par réseaux de neurones et Support Vector Machine". Toulouse 2, 2005. http://www.theses.fr/2005TOU20089.
Texto completoIn this thesis, we first present the results of an interdisciplinary project in which we use the approximation abilities of multilayer perceptrons in order to predict land cover maps. Subsequently, we focus on the extension of the neural networks and of the SVM for functional data analysis. Our purpose is to build non linear tools for functional data. A part of our work is based on a semi-parametric approach which uses a functional inverse regression method. Then, we present another approach which allows us to build kernels for SVM in order to take into account the functional nature of the data. In this work, the statistical learning theory plays a central role and we apply ourselves to give consistency results for our methods, as much as possible
Azeraf, Elie. "Classification avec des modèles probabilistes génératifs et des réseaux de neurones. Applications au traitement des langues naturelles". Electronic Thesis or Diss., Institut polytechnique de Paris, 2022. https://theses.hal.science/tel-03880848.
Texto completoMany probabilistic models have been neglected for classification tasks with supervised learning for several years, as the Naive Bayes or the Hidden Markov Chain. These models, called generative, are criticized because the induced classifier must learn the observations' law. This problem is too complex when the number of observations' features is too large. It is especially the case with Natural Language Processing tasks, as the recent embedding algorithms convert words in large numerical vectors to achieve better scores.This thesis shows that every generative model can define its induced classifier without using the observations' law. This proposition questions the usual categorization of the probabilistic models and classifiers and allows many new applications. Therefore, Hidden Markov Chain can be efficiently applied to Chunking and Naive Bayes to sentiment analysis.We go further, as this proposition allows to define the classifier induced from a generative model with neural network functions. We "neuralize" the models mentioned above and many of their extensions. Models so obtained allow to achieve relevant scores for many Natural Language Processing tasks while being interpretable, able to require little training data, and easy to serve
Martin, Philippe. "Réseaux de neurones artificiels : application à la reconnaissance optique de partitions musicales". Phd thesis, Grenoble 1, 1992. http://tel.archives-ouvertes.fr/tel-00340938.
Texto completoVasilache, Adriana. "Les réseaux de neurones pour la modélisation et la commande des procédés biotechnologiques". Toulouse, INSA, 2000. http://www.theses.fr/2000ISAT0050.
Texto completoIn this work we realize a study on the use of the neural nets for the modeling, classification and the control of fermentation processes. The black-box models (we consider a neural net like a black box model) are of great help for processes or phenomena modeling when analytical models cannot be deduced from physical considerations. Some of the advantages of the neural nets when compared to other black-box models are: they are universal approximators using a small number of parameters, their basis functions are adaptive, their repetitive structure permits an easy implementation both software and hardware and they have the property of implicit regularization. These, combined with the characteristics of the biological processes (which are non-linear, non-stationary processes whose dynamics isn’t entirely known), are the reason for which the neural nets are used for the modeling of such processes. We have thus used existing neural models and proposed new ones for the cases of lactic and alcoholic fermentations. We have presented two approaches for the characterization of the fermentation process dynamics: the modeling of the specific biomass growth rate, the most important dynamic parameter of a fermentation process and the global characterization of the process dynamics using a neural classifier. The two approaches have been tested in simulation and on real data for lactic or alcoholic fermentation processes. The use of a classifier of the process dynamics represents a potential tool for process supervision by means of detecting the changes in the process dynamics as well as an aid for the process modeling in the case of batch processes. The prediction of the biomass concentration has also been considered for a continuous fermentation process. The neural models have been tested in a predictive control strategy and compared with a similar strategy using adaptive modeling. The neural prediction has been an incontestable winner for the cases where the process dynamics changes in time. The last issue of our study has been the prediction of the respiratory quotient for a alcoholic fermentation for which we proposed a neural model. It has been proposed in view of a predictive control strategy for the maintenance of a certain regime (fermentative or oxidative)
Guerry, Joris. "Reconnaissance visuelle robuste par réseaux de neurones dans des scénarios d'exploration robotique. Détecte-moi si tu peux !" Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLX080/document.
Texto completoThe main objective of this thesis is visual recognition for a mobile robot in difficult conditions. We are particularly interested in neural networks which present today the best performances in computer vision. We studied the concept of method selection for the classification of 2D images by using a neural network selector to choose the best available classifier given the observed situation. This strategy works when data can be easily partitioned with respect to available classifiers, which is the case when complementary modalities are used. We have therefore used RGB-D data (2.5D) in particular applied to people detection. We propose a combination of independent neural network detectors specific to each modality (color & depth map) based on the same architecture (Faster RCNN). We share intermediate results of the detectors to allow them to complement and improve overall performance in difficult situations (luminosity loss or acquisition noise of the depth map). We are establishing new state of the art scores in the field and propose a more complex and richer data set to the community (ONERA.ROOM). Finally, we made use of the 3D information contained in the RGB-D images through a multi-view method. We have defined a strategy for generating 2D virtual views that are consistent with the 3D structure. For a semantic segmentation task, this approach artificially increases the training data for each RGB-D image and accumulates different predictions during the test. We obtain new reference results on the SUNRGBD and NYUDv2 datasets. All these works allowed us to handle in an original way 2D, 2.5D and 3D robotic data with neural networks. Whether for classification, detection and semantic segmentation, we not only validated our approaches on difficult data sets, but also brought the state of the art to a new level of performance
Attik, Mohammed. "Traitement intelligent de données par réseaux de neurones artificiels : application à la valorisation des systèmes d'information géographiques". Nancy 1, 2006. http://docnum.univ-lorraine.fr/public/SCD_T_2006_0211_ATTIK.pdf.
Texto completoThe purpose of this thesis is: (i) establish predictive maps on ore deposits, (ii) select a subset of descriptive features that effectively contribute to the building of these predictive maps, (iii) identify and interpret dependencies between the selected features, (iv) place the features into a hierarchy that indicates their importance. A real-life data of Geographical Information System provided by the French geological survey (BRGM) have been used in the accomplished experiments. In order to establish predictive maps, we have used neural network ensemble which is a very successful technique where outputs of a set of separately trained neural network are combined to form one unified prediction. This technique generates several predictive maps following the used aggregation function. In addition, to understand domain data, we have focused on selecting a subset of relevant features. We have proposed an improvement of existing features selection techniques that are based on the principle of Optimal Brain Damage (OBD) as well as those of Optimal Brain Surgeon (OBS) and Mutual Information (MI). We have also proposed novel solutions to understand data that combine ensemble feature selection approach with either concept lattices or statistic techniques. The latter solutions help discovering all relevant features and organizing them into hierarchy according to their concurrencies in the selected subsets of features. Moreover, we have addressed the problem of clustering-based analysis of data provided with multiple labels. The proposed approach uses new measures that extend the scope of the recall and precision measures in information retrieval (IR) to the processing of multi-label data. Experiments have been carried out on data pertaining to geographical information system and documentary system have highlighted the accuracy of our approach for knowledge extraction
Ziat, Ali Yazid. "Apprentissage de représentation pour la prédiction et la classification de séries temporelles". Thesis, Paris 6, 2017. http://www.theses.fr/2017PA066324/document.
Texto completoThis thesis deals with the development of time series analysis methods. Our contributions focus on two tasks: time series forecasting and classification. Our first contribution presents a method of prediction and completion of multivariate and relational time series. The aim is to be able to simultaneously predict the evolution of a group of time series connected to each other according to a graph, as well as to complete the missing values in these series (which may correspond for example to a failure of a sensor during a given time interval). We propose to use representation learning techniques to forecast the evolution of the series while completing the missing values and taking into account the relationships that may exist between them. Extensions of this model are proposed and described: first in the context of the prediction of heterogeneous time series and then in the case of the prediction of time series with an expressed uncertainty. A prediction model of spatio-temporal series is then proposed, in which the relations between the different series can be expressed more generally, and where these can be learned.Finally, we are interested in the classification of time series. A joint model of metric learning and time-series classification is proposed and an experimental comparison is conducted
Martin, Pierre-Etienne. "Détection et classification fines d'actions à partir de vidéos par réseaux de neurones à convolutions spatio-temporelles : Application au tennis de table". Thesis, Bordeaux, 2020. http://www.theses.fr/2020BORD0313.
Texto completoAction recognition in videos is one of the key problems in visual data interpretation. Despite intensive research, differencing and recognizing similar actions remains a challenge. This thesis deals with fine-grained classification of sport gestures from videos, with an application to table tennis.In this manuscript, we propose a method based on deep learning for automatically segmenting and classifying table tennis strokes in videos. Our aim is to design a smart system for students and teachers for analyzing their performances. By profiling the players, a teacher can therefore tailor the training sessions more efficiently in order to improve their skills. Players can also have an instant feedback on their performances.For developing such a system with fine-grained classification, a very specific dataset is needed to supervise the learning process. To that aim, we built the “TTStroke-21” dataset, which is composed of 20 stroke classes plus a rejection class. The TTStroke-21 dataset comprises video clips of recorded table tennis exercises performed by students at the sport faculty of the University of Bordeaux - STAPS. These recorded sessions were annotated by professional players or teachers using a crowdsourced annotation platform. The annotations consist in a description of the handedness of the player and information for each stroke performed (starting and ending frames, class of the stroke).Fine-grained action recognition has some notable differences with coarse-grained action recognition. In general, datasets used for coarse-grained action recognition, the background context often provides discriminative information that methods can use to classify the action, rather than focusing on the action itself. In fine-grained classification, where the inter-class similarity is high, discriminative visual features are harder to extract and the motion plays a key role for characterizing an action.In this thesis, we introduce a Twin Spatio-Temporal Convolutional Neural Network. This deep learning network takes as inputs an RGB image sequence and its computed Optical Flow. The RGB image sequence allows our model to capture appearance features while the optical flow captures motion features. Those two streams are processed in parallel using 3D convolutions, and fused at the last stage of the network. Spatio-temporal features extracted in the network allow efficient classification of video clips from TTStroke-21. Our method gets an average classification performance of 87.3% with a best run of 93.2% accuracy on the test set. When applied on joint detection and classification task, the proposed method reaches an accuracy of 82.6%.A systematic study of the influence of each stream and fusion types on classification accuracy has been performed, giving clues on how to obtain the best performances. A comparison of different optical flow methods and the role of their normalization on the classification score is also done. The extracted features are also analyzed by back-tracing strong features from the last convolutional layer to understand the decision path of the trained model. Finally, we introduce an attention mechanism to help the model focusing on particular characteristic features and also to speed up the training process. For comparison purposes, we provide performances of other methods on TTStroke-21 and test our model on other datasets. We notice that models performing well on coarse-grained action datasets do not always perform well on our fine-grained action dataset.The research presented in this manuscript was validated with publications in one international journal, five international conference papers, two international workshop papers and a reconductible task in MediaEval workshop in which participants can apply their action recognition methods to TTStroke-21. Two additional international workshop papers are in process along with one book chapter
Ziat, Ali Yazid. "Apprentissage de représentation pour la prédiction et la classification de séries temporelles". Electronic Thesis or Diss., Paris 6, 2017. http://www.theses.fr/2017PA066324.
Texto completoThis thesis deals with the development of time series analysis methods. Our contributions focus on two tasks: time series forecasting and classification. Our first contribution presents a method of prediction and completion of multivariate and relational time series. The aim is to be able to simultaneously predict the evolution of a group of time series connected to each other according to a graph, as well as to complete the missing values in these series (which may correspond for example to a failure of a sensor during a given time interval). We propose to use representation learning techniques to forecast the evolution of the series while completing the missing values and taking into account the relationships that may exist between them. Extensions of this model are proposed and described: first in the context of the prediction of heterogeneous time series and then in the case of the prediction of time series with an expressed uncertainty. A prediction model of spatio-temporal series is then proposed, in which the relations between the different series can be expressed more generally, and where these can be learned.Finally, we are interested in the classification of time series. A joint model of metric learning and time-series classification is proposed and an experimental comparison is conducted
Ibbou, Smaïl. "Classification, analyse des correspondances et methodes neuronales". Paris 1, 1998. http://www.theses.fr/1998PA010020.
Texto completoFirmin, Christian. "Optimisation des réseaux de neurones à fonctions radiales de base par critères informationnels : application à la détection de défauts en production de bouteilles". Lille 1, 1997. http://www.theses.fr/1997LIL10048.
Texto completoBrooks, Daniel. "Deep Learning and Information Geometry for Time-Series Classification". Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS276.
Texto completoMachine Learning, and in particular Deep Learning, is a powerful tool to model and study the intrinsic statistical foundations of data, allowing the extraction of meaningful, human-interpretable information from otherwise unpalatable arrays of floating points. While it provides a generic solution to many problems, some particular data types exhibit strong underlying physical structure: images have spatial locality, audio has temporal sequentiality, radar has time-frequency structure. Both intuitively and formally, there can be much to gain in leveraging this structure by adapting the subsequent learning models. As convolutional architectures for images, signal properties can be encoded and harnessed within the network. Conceptually, this would allow for a more intrinsic handling of the data, potentially leading to more efficient learning models. Thus, we will aim to use known structures in the signals as model priors. Specifically, we build dedicated deep temporal architectures for time series classification, and explore the use of complex values in neural networks to further refine the analysis of structured data. Going even further, one may wish to directly study the signal’s underlying statistical process. As such, Gaussian families constitute a popular candidate. Formally, the covariance of the data fully characterizes such a distribution; developing Machine Learning algorithms on covariance matrices will thus be a central theme throughout this thesis. Statistical distributions inherently diverge from the Euclidean framework; as such, it is necessary to study them on the appropriate, curved Riemannian manifold, as opposed to a flat, Euclidean space. Specifically, we contribute to existing deep architectures by adding normalizations in the form of data-aware mappings, and a Riemannian Batch Normalization algorithm. We showcase empirical validation through a variety of different tasks, including emotion and action recognition from video and Motion Capture data, with a sharpened focus on micro-Doppler radar data for Non-Cooperative Target Recognition drone recognition. Finally, we develop a library for the Deep Learning framework PyTorch, to spur reproducibility and ease of use
Rouzier, Sophie. "Réseaux neuronaux et modularité". Grenoble INPG, 1998. http://www.theses.fr/1998INPG0032.
Texto completoBriquet-Kerestedjian, Nolwenn. "Impact detection and classification for safe physical Human-Robot Interaction under uncertainties". Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLC038/document.
Texto completoThe present thesis aims to develop an efficient strategy for impact detection and classification in the presence of modeling uncertainties of the robot and its environment and using a minimum number of sensors, in particular in the absence of force/torque sensor.The first part of the thesis deals with the detection of an impact that can occur at any location along the robot arm and at any moment during the robot trajectory. Impact detection methods are commonly based on a dynamic model of the system, making them subject to the trade-off between sensitivity of detection and robustness to modeling uncertainties. In this respect, a quantitative methodology has first been developed to make explicit the contribution of the errors induced by model uncertainties. This methodology has been applied to various detection strategies, based either on a direct estimate of the external torque or using disturbance observers, in the perfectly rigid case or in the elastic-joint case. A comparison of the type and structure of the errors involved and their consequences on the impact detection has been deduced. In a second step, novel impact detection strategies have been designed: the dynamic effects of the impacts are isolated by determining the maximal error range due to modeling uncertainties using a stochastic approach.Once the impact has been detected and in order to trigger the most appropriate post-impact robot reaction, the second part of the thesis focuses on the classification step. In particular, the distinction between an intentional contact (the human operator intentionally interacts with the robot, for example to reconfigure the task) and an undesired contact (a human subject accidentally runs into the robot), as well as the localization of the contact on the robot, is investigated using supervised learning techniques and more specifically feedforward neural networks. The challenge of generalizing to several human subjects and robot trajectories has been investigated
Cabanes, Emmanuel. "Traitement et analyse des données acquises par spectrométrie de résonance magnétique in vivo : Evaluation de la méthode HSLVD et de la classification par réseaux de neurones". Aix-Marseille 2, 2001. http://www.theses.fr/2001AIX22008.
Texto completoVodenicarevic, Damir. "Rhythms and oscillations : a vision for nanoelectronics". Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLS518/document.
Texto completoWith the advent of "artificial intelligence", computers, mobile devices and other connected objects are being pushed beyond the realm of arithmetic and logic operations, for which they have been optimized over decades, in order to process "cognitive" tasks such as automatic translation and image or voice recognition, for which they are not the ideal substrate. As a result, supercomputers may require megawatts to process tasks for which the human brain only needs 20 watt. This has revived interest into the design of alternative computing schemes inspired by the brain. In particular, neural oscillations that appear to be linked to computational activity in the brain have inspired approaches leveraging the complex physics of networks of coupled oscillators in order to process cognitive tasks efficiently. In the light of recent advances in nano-technology allowing the fabrication of highly integrable nano-oscillators, this thesis proposes and studies novel neuro-inspired oscillator-based pattern classification architectures that could be implemented on chip
Ben, Mustapha Zied. "Télédétection des groupes phytoplanctoniques via l'utilisation conjointe de mesures satellites, in situ et d'une méthode de classification automatique". Thesis, Littoral, 2013. http://www.theses.fr/2013DUNK0405/document.
Texto completoRemote sensing of ocean color is a powerful tool for monitoring phytoplankton in the ocean with a high spatial and temporal resolution. Several methods were developed in the past years for detecting phytoplankton functional types from satellite observations. In this thesis, we present an automatic classification method, based on a neural network clustering algorithm, in order to classify the anomalies of water leaving radiances spectra (Ra), introduced in the PHYSAT method by Alvain et al. (2005) and analyze their variability at the global scale. The use of an unsupervised classification aims at improving the characterization of the spectral variability of Ra in terms of shape and amplitude as well as the expansion of its potential use to larger in situ datasets for global phytoplankton remote sensing. The Self-Organizing Map Algorithm (SOM) aggregates similar spectra into a reduced set of pertinent groups, allowing the characterization of the Ra variability, which is known to be linked with the phytoplankton community composition. Based on the same sample of Ra spectra, a comparison between the previous version of PHYSAT and the new one using SOM shows that is now possible to take into consideration all the types of spectra. This was not possible with the previous approach, based on thresholds, defined in order to avoid overlaps between the spectral signatures of each phytoplankton group. The SOM-based method is relevant for characterizing a wide variety of Ra spectra through its ability to handle large amounts of data, in addition to its statistical reliability compared to the previous PHYSAT. The former approach might have introduced potential biases and thus, its extension to larger databases was very restricted. In a second step, some new Ra spectra have been related to phytoplankton groups using collocated field pigments inventories from a large in situ database. Phytoplankton groups were identified based on biomarker pigments ratios thresholds taken from the literature. SOM was then applied to the global daily SeaWiFS imagery archive between 1997 and 2010. Global distributions of major phytoplankton groups were analyzed and validated against in situ data. Thanks to its ability to capture a wide range of spectra and to manage a larger in situ pigment dataset, the neural network tool allows to classify a much higher number of pixels (2 times more) than the previous PHYSAT method for the five phytoplankton groups taken into account in this study (Synechococcus-Like-Cyanobacteria, diatoms, Prochloroccus, Nanoeucaryots and Phaeocystis-like). In addition, different Ra spectral signatures have been associated to diatoms. These signatures are located in various environments where the inherent optical properties affecting the Ra spectra are likely to be significantly different. Local phenomena such as diatoms blooms in the upwelling regions or during climatic events(i.e. La Nina) are more clearly visible with the new method. The PHYSAT-SOM method provides several perspectives concerning the use of the ocean color remote sensing data for phytoplankton group identification, such as, the potential application of the method in Case 2 waters, using an appropriate nLw signal normalization approach. A preliminary case study in the English Channel and North Sea waters is presented in the last chapter of the thesis, showing the possibility of a future use of PHYSAT-SOM in these optically complex waters
Chiaroni, Florent. "Weakly supervised learning for image classification and potentially moving obstacles analysis". Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASC006.
Texto completoIn the context of autonomous vehicle perception, the interest of the research community for deep learning approaches has continuously grown since the last decade. This can be explained by the fact that deep learning techniques provide nowadays state-of-the-art prediction performances for several computer vision challenges. More specifically, deep learning techniques can provide rich semantic information concerning the complex visual patterns encountered in autonomous driving scenarios. However, such approaches require, as their name implies, to learn on data. In particular, state-of-the-art prediction performances on discriminative tasks often demand hand labeled data of the target application domain. Hand labeling has a significant cost, while, conversely, unlabeled data can be easily obtained in the autonomous driving context. It turns out that a category of learning strategies, referred to as weakly supervised learning, enables to exploit partially labeled data. Therefore, we aim in this thesis at reducing as much as possible the hand labeling requirement by proposing weakly supervised learning techniques.We start by presenting a type of learning methods which are self-supervised. They consist of substituting hand-labels by upstream techniques able to automatically generate exploitable training labels. Self-supervised learning (SSL) techniques have proven their usefulness in the past for offroad obstacles avoidance and path planning through changing environments. However, SSL techniques still leave the door open for detection, segmentation, and classification of static potentially moving obstacles.Consequently, we propose in this thesis three novel weakly supervised learning methods with the final goal to deal with such road users through an SSL framework. The first two proposed contributions of this work aim at dealing with partially labeled image classification datasets, such that the labeling effort can be only focused on our class of interest, the positive class. Then, we propose an approach which deals with training data containing a high fraction of wrong labels, referred to as noisy labels. Next, we demonstrate the potential of such weakly supervised strategies for detection and segmentation of potentially moving obstacles
Sayadi, Karim. "Classification du texte numérique et numérisé. Approche fondée sur les algorithmes d'apprentissage automatique". Thesis, Paris 6, 2017. http://www.theses.fr/2017PA066079/document.
Texto completoDifferent disciplines in the humanities, such as philology or palaeography, face complex and time-consuming tasks whenever it comes to examining the data sources. The introduction of computational approaches in humanities makes it possible to address issues such as semantic analysis and systematic archiving. The conceptual models developed are based on algorithms that are later hard coded in order to automate these tedious tasks. In the first part of the thesis we propose a novel method to build a semantic space based on topics modeling. In the second part and in order to classify historical documents according to their script. We propose a novel representation learning method based on stacking convolutional auto-encoder. The goal is to automatically learn plot representations of the script or the written language
Cholet, Stéphane. "Evaluation automatique des états émotionnels et dépressifs : vers un système de prévention des risques psychosociaux". Thesis, Antilles, 2019. http://www.theses.fr/2019ANTI0388/document.
Texto completoPsychosocial risks are a major public health issue, because of the disorders they can trigger : stress, mood swings, burn-outs, etc. Although propoer diagnosis can only be made by a healthcare professionnel, Affective Computing can make a contribution by improving the understanding of the phenomena. Affective Computing is a multidisciplinary field involving concepts of Artificial Intelligence, psychology and psychiatry, among others. In this research, we are interested in two elements that can be subject to disorders: the emotional state and the depressive state of individuals.The concept of emotion covers a wide range of definitions and models, most of which are based on work in psychiatry or psychology. A famous example is Russell's circumplex, which defines an emotion as the combination of two emotional dimensions, called valence and arousal. Valence denotes an individual's sad or joyful character, while arousal denotes his passive or active character. The automatic evaluation of emotional states has generated a significant revival of interest in the last decade. Methods from Artificial Intelligence allow to achieve interesting performances, from data captured in a non-invasive manner, such as videos. However, there is one aspect that has not been studied much: that of emotional intensities and the possibility of recognizing them. In this thesis, we have explored this aspect using visualization and classification methods to show that the use of emotional intensity classes, rather than continuous values, benefits both automatic recognition and state interpretation.The concept of depression is more strict, as it is a recognized disease as such. It affects individuals regardless of age, gender or occupation, but varies in intensity or nature of symptoms. For this reason, its study, both at the level of detection and monitoring, is of major interest for the prevention of psychosocial risks.However, his diagnosis is made difficult by the sometimes innocuous nature of the symptoms and by the often delicate process of consulting a specialist. The Beck's scale and the associated score allow, by means of a questionnaire, to evaluate the severity of an individual's state of depression. The system we have developed is able to automatically recognize an individual's depressive score from videos. It includes, on the one hand, a low-level visual spatio-temporal descriptor that quantifies micro and macro facial movements and, on the other hand, neural methods from the cognitive sciences. Its speed allows applications for real-time recognition of depressive states, and its performance is interesting with regard to the state of the art. The fusion of visual and auditory modalities has also been studied, showing that the use of these two sensory channels benefits the recognition of depressive states.Beyond performance and originality, one of the strong points of this thesis is the interpretability of the methods. Indeed, in a multidisciplinary context such as that of Affective Computing, improving knowledge and understanding of the studied phenomena is a key point that usual computer methods implemeted as "black boxes" can't deal with
Anjum, Ayesha. "Differentiation of alzheimer's disease dementia, mild cognitive impairment and normal condition using PET-FDG and AV-45 imaging : a machine-learning approach". Toulouse 3, 2013. http://thesesups.ups-tlse.fr/2238/.
Texto completoWe used PET imaging with tracers F18-FDG and AV45 in conjunction with the classification methods in the field of "Machine Learning". PET images were acquired in dynamic mode, an image every 5 minutes. The images used come from three different sources: the database ADNI (Alzheimer's Disease Neuro-Imaging Initiative, University of California Los Angeles) and two protocols performed in the PET center of the Purpan Hospital. The classification was applied after processing dynamic images by Principal Component Analysis and Independent Component Analysis. The data were separated into training set and test set. To evaluate the performance of the classification we used the method of cross-validation LOOCV (Leave One Out Cross Validation). We give a comparison between the two most widely used classification methods, SVM (Support Vector Machine) and artificial neural networks (ANN) for both tracers. The combination giving the best classification rate seems to be SVM and AV45 tracer. However the most important confusion is found between MCI patients and normal subjects. Alzheimer's patients differ somewhat better since they are often found in more than 90%. We evaluated the generalization of our methods by making learning from set of data and classification on another set. We reached the specifity score of 100% and sensitivity score of more than 81%. SVM method showed a bettrer sensitivity than Artificial Neural Network method. The value of such work is to help the clinicians in diagnosing Alzheimer's disease
Parfait, Sébastien. "Classification de spectres et recherche de biomarqueurs en spectroscopie par résonance magnétique nucléaire du proton dans les tumeurs prostatiques". Phd thesis, Université de Bourgogne, 2010. http://tel.archives-ouvertes.fr/tel-00596568.
Texto completoRemm, Jean-François. "Extraction de connaissances par réseaux neuronaux : application au domaine du radar". Nancy 1, 1996. http://www.theses.fr/1996NAN10366.
Texto completoBouju, Alain. "Etiquetage et poursuite de points caractéristiques d'un objet 3D par des méthodes connexionistes". Toulouse, ENSAE, 1993. http://www.theses.fr/1993ESAE0017.
Texto completoBailly, Adeline. "Classification de séries temporelles avec applications en télédétection". Thesis, Rennes 2, 2018. http://www.theses.fr/2018REN20021/document.
Texto completoTime Series Classification (TSC) has received an important amount of interest over the past years due to many real-life applications. In this PhD, we create new algorithms for TSC, with a particular emphasis on Remote Sensing (RS) time series data. We first propose the Dense Bag-of-Temporal-SIFT-Words (D-BoTSW) method that uses dense local features based on SIFT features for 1D data. Extensive experiments exhibit that D-BoTSW significantly outperforms nearly all compared standalone baseline classifiers. Then, we propose an enhancement of the Learning Time Series Shapelets (LTS) algorithm called Adversarially-Built Shapelets (ABS) based on the introduction of adversarial time series during the learning process. Adversarial time series provide an additional regularization benefit for the shapelets and experiments show a performance improvementbetween the baseline and our proposed framework. Due to the lack of available RS time series datasets,we also present and experiment on two remote sensing time series datasets called TiSeLaCand Brazilian-Amazon
Haj, Hassan Hawraa. "Détection et classification temps réel de biocellules anormales par technique de segmentation d’images". Thesis, Université de Lorraine, 2018. http://www.theses.fr/2018LORR0043.
Texto completoDevelopment of methods for help diagnosis of the real time detection of abnormal cells (which can be considered as cancer cells) through bio-image processing and detection are most important research directions in information science and technology. Our work has been concerned by developing automatic reading procedures of the normal and abnormal bio-images tissues. Therefore, the first step of our work is to detect a certain type of abnormal bio-images associated to many types evolution of cancer within a Microscopic multispectral image, which is an image, repeated in many wavelengths. And using a new segmentation method that reforms itself in an iterative adaptive way to localize and cover the real cell contour, using some segmentation techniques. It is based on color intensity and can be applied on sequences of objects in the image. This work presents a classification of the abnormal tissues using the Convolution neural network (CNN), where it was applied on the microscopic images segmented using the snake method, which gives a high performance result with respect to the other segmentation methods. This classification method reaches high performance values, where it reaches 100% for training and 99.168% for testing. This method was compared to different papers that uses different feature extraction, and proved its high performance with respect to other methods. As a future work, we will aim to validate our approach on a larger datasets, and to explore different CNN architectures and the optimization of the hyper-parameters, in order to increase its performance, and it will be applied to relevant medical imaging tasks including computer-aided diagnosis
Ben, Naceur Mostefa. "Deep Neural Networks for the segmentation and classification in Medical Imaging". Thesis, Paris Est, 2020. http://www.theses.fr/2020PESC2014.
Texto completoNowadays, getting an efficient segmentation of Glioblastoma Multiforme (GBM) braintumors in multi-sequence MRI images as soon as possible, gives an early clinical diagnosis, treatment, and follow-up. The MRI technique is designed specifically to provide radiologists with powerful visualization tools to analyze medical images, but the challenge lies more in the information interpretation of radiological images with clinical and pathologies data and their causes in the GBM tumors. This is why quantitative research in neuroimaging often requires anatomical segmentation of the human brain from MRI images for the detection and segmentation of brain tumors. The objective of the thesis is to propose automatic Deep Learning methods for brain tumors segmentation using MRI images.First, we are mainly interested in the segmentation of patients’ MRI images with GBMbrain tumors using Deep Learning methods, in particular, Deep Convolutional NeuralNetworks (DCNN). We propose two end-to-end DCNN-based approaches for fully automaticbrain tumor segmentation. The first approach is based on the pixel-wise techniquewhile the second one is based on the patch-wise technique. Then, we prove that thelatter is more efficient in terms of segmentation performance and computational benefits. We also propose a new guided optimization algorithm to optimize the suitable hyperparameters for the first approach. Second, to enhance the segmentation performance of the proposed approaches, we propose new segmentation pipelines of patients’ MRI images, where these pipelines are based on deep learned features and two stages of training. We also address problems related to unbalanced data in addition to false positives and false negatives to increase the model segmentation sensitivity towards the tumor regions and specificity towards the healthy regions. Finally, the segmentation performance and the inference time of the proposed approaches and pipelines are reported along with state-of-the-art methods on a public dataset annotated by radiologists and approved by neuroradiologists
Oukhellou, Latifa. "Paramétrisation et classification de signaux en contrôle non destructif. Application à la reconnaissance des défauts de rails par courants de Foucault". Phd thesis, Université Paris Sud - Paris XI, 1997. http://tel.archives-ouvertes.fr/tel-00006600.
Texto completoHerry, Sébastien. "Détection automatique de langue par discrimination d'experts". Paris 6, 2007. http://www.theses.fr/2007PA066101.
Texto completoThe purpose of the presented work in this memoir is to automatically detect language in audio stream. For this we suggest a model which, like bilingual expert, done an discrimination by language pair with only acoustic information. The system have constraint : Operating in real time, Use database without phonetic information, Able to add a new language without retrain all the model In a first time we have done an Automatic language detection system derived from the stat of the art. The results obtained by this system are used as reference for the rest of memoir, and we compare those results with the results obtained by the developed model. In a first time, we propose a set of discriminator, by pair of language, based on neural network. The treatment is done on the whole duration of speech segment. The results of these discriminators are fused to create de detection. This model has a patent. We have study more precisely the influence of different parameter as the number of locator, the variation intra and inter corpus or the hardiness. Next we have compared the proposed modelling based on discrimination, with modelling auto regressive or predictive. This system has been tested with our participation of the international campaign organised by NIST in December 2005. To conclude on this campaign where 17 international teams have participated, we have proposed several improvements as: A normalisation of database, A modification of speaker database for learning only, Increase scores with segment duration. To conclude, the system proposed fulfils the constraints because the system is real time, and use only acoustic information. More over the system is more efficient than the derived model from the stat of the art. At last the model is hardiness for noise, for unknown language, for new evaluation database
Tomasini, Linda. "Apprentissage d'une représentation statistique et topologique d'un environnement". Toulouse, ENSAE, 1993. http://www.theses.fr/1993ESAE0024.
Texto completoBellanger-Dujardin, Anne-Sophie. "Contribution à l'étude de structures neuronales pour la classification de signatures : application au diagnostic de pannes des systèmes industriels et à l'aide au diagnostic médical". Paris 12, 2003. https://athena.u-pec.fr/primo-explore/search?query=any,exact,990002111250204611&vid=upec.
Texto completoThe problem of diagnosis occurs in many fields, especially medical and industrial, where operator has a key role. The major difficulty bound to this problem lies on the resemblance between the signatures which allow to make a diagnosis. Furthermore, we often have an empirical knowledge of the system, and thus, an incomplete model, requiring the appeal to an expert. Our efforts were focused on techniques based on neural techniques for computer aided diagnosis. For the tasks of pattern recognition, classification and decision, the proposed techniques indeed presents a number of advantages over conventional models because of their abilities of learning and generalization. Moreover, noticing that simple neural techniques do not allow obtaining good results, we propose a neural hybrid structure. Two areas of applications have been considered: one linked to the biomedical field and the other concerning the industrial domain
Monrousseau, Thomas. "Développement du système d'analyse des données recueillies par les capteurs et choix du groupement de capteurs optimal pour le suivi de la cuisson des aliments dans un four". Thesis, Toulouse, INSA, 2016. http://www.theses.fr/2016ISAT0054.
Texto completoIn a world where all personal devices become smart and connected, some French industrials created a project to make ovens able detecting the cooking state of fish and meat without contact sensor. This thesis takes place in this context and is divided in two major parts. The first one is a feature selection phase to be able to classify food in three states: under baked, well baked and over baked. The point of this selection method, based on fuzzy logic is to strongly reduce the number of features got from laboratory specific sensors. The second part concerns on-line monitoring of the food cooking state by several methods. These technics are: classification algorithm into ten bake states, the use of a discrete version of the heat equation and the development of a soft sensor based on an artificial neural network model build from cooking experiments to infer the temperature inside the food from available on-line measurements. These algorithms have been implemented on microcontroller equipping a prototype version of a new oven in order to be tested and validated on real use cases
Durand, Stéphane. "TOM, une architecture connexionniste de traitement de séquences : application à la reconnaissance de la parole". Nancy 1, 1995. http://www.theses.fr/1995NAN10411.
Texto completoLéon, Aurélia. "Apprentissage séquentiel budgétisé pour la classification extrême et la découverte de hiérarchie en apprentissage par renforcement". Electronic Thesis or Diss., Sorbonne université, 2019. http://www.theses.fr/2019SORUS226.
Texto completoThis thesis deals with the notion of budget to study problems of complexity (it can be computational complexity, a complex task for an agent, or complexity due to a small amount of data). Indeed, the main goal of current techniques in machine learning is usually to obtain the best accuracy, without worrying about the cost of the task. The concept of budget makes it possible to take into account this parameter while maintaining good performances. We first focus on classification problems with a large number of classes: the complexity in those algorithms can be reduced thanks to the use of decision trees (here learned through budgeted reinforcement learning techniques) or the association of each class with a (binary) code. We then deal with reinforcement learning problems and the discovery of a hierarchy that breaks down a (complex) task into simpler tasks to facilitate learning and generalization. Here, this discovery is done by reducing the cognitive effort of the agent (considered in this work as equivalent to the use of an additional observation). Finally, we address problems of understanding and generating instructions in natural language, where data are available in small quantities: we test for this purpose the simultaneous use of an agent that understands and of an agent that generates the instructions
Sampaio, de Rezende Rafael. "New methods for image classification, image retrieval and semantic correspondence". Thesis, Paris Sciences et Lettres (ComUE), 2017. http://www.theses.fr/2017PSLEE068/document.
Texto completoThe problem of image representation is at the heart of computer vision. The choice of feature extracted of an image changes according to the task we want to study. Large image retrieval databases demand a compressed global vector representing each image, whereas a semantic segmentation problem requires a clustering map of its pixels. The techniques of machine learning are the main tool used for the construction of these representations. In this manuscript, we address the learning of visual features for three distinct problems: Image retrieval, semantic correspondence and image classification. First, we study the dependency of a Fisher vector representation on the Gaussian mixture model used as its codewords. We introduce the use of multiple Gaussian mixture models for different backgrounds, e.g. different scene categories, and analyze the performance of these representations for object classification and the impact of scene category as a latent variable. Our second approach proposes an extension to the exemplar SVM feature encoding pipeline. We first show that, by replacing the hinge loss by the square loss in the ESVM cost function, similar results in image retrieval can be obtained at a fraction of the computational cost. We call this model square-loss exemplar machine, or SLEM. Secondly, we introduce a kernelized SLEM variant which benefits from the same computational advantages but displays improved performance. We present experiments that establish the performance and efficiency of our methods using a large array of base feature representations and standard image retrieval datasets. Finally, we propose a deep neural network for the problem of establishing semantic correspondence. We employ object proposal boxes as elements for matching and construct an architecture that simultaneously learns the appearance representation and geometric consistency. We propose new geometrical consistency scores tailored to the neural network’s architecture. Our model is trained on image pairs obtained from keypoints of a benchmark dataset and evaluated on several standard datasets, outperforming both recent deep learning architectures and previous methods based on hand-crafted features. We conclude the thesis by highlighting our contributions and suggesting possible future research directions
Bertrand, Sarah. "Analyse d'images pour l'identification multi-organes d'espèces végétales". Thesis, Lyon, 2018. http://www.theses.fr/2018LYSE2127/document.
Texto completoThis thesis is part of the ANR ReVeRIES, which aims to use mobile technologies to help people better understand their environment and in particular the plants that surround them. More precisely, the ReVeRIES project is based on a mobile application called Folia developed as part of the ANR ReVeS project and capable of recognising tree and shrub species based on photos of their leaves. This prototype differs from other tools in that it is able to simulate the behaviour of the botanist. In the context of the ReVeRIES project, we propose to go much further by developing new aspects: multimodal species recognition, learning through play and citizen science. The purpose of this thesis is to focus on the first of these three aspects, namelythe analysis of images of plant organs for identification.More precisely, we consider the main trees and shrubs, endemic or exotic, found in metropolitan France. The objective of this thesis is to extend the recognition algorithm by taking into account other organs in addition to the leaf. This multi-modality is indeed essential if we want the user to learn and practice the different methods of recognition for which botanists use the variety of organs (i.e. leaves, flowers, fruits and bark). The method used by Folia for leaf recognition being dedicated, because simulating the work of a botanist on the leaf, cannot be applied directly to other organs. Thus, new challenges are emerging, both in terms of image processing and data fusion.The first part of the thesis was devoted to the implementation of image processing methods for the identification of plant species. The identification of tree species from bark images was the first to be studied. The descriptors developed take into account the structure of the bark inspired from the criteria used by botanists. Fruits and flowers required a segmentation step before their description. A new segmentation method that can be used on smartphones has been developed to work in spite of the high variability of flowers and fruits. Finally, descriptors were extracted on fruits and flowers after the segmentation step. We decided not to separate flowers and fruits because we showed that a user new to botany does not always know the difference between these two organs on so-called "ornamental" trees (not fruit trees). For fruits and flowers, prediction is not only made on their species but also on their genus and family, botanical groups reflecting a similarity between these organs.The second part of the thesis deals with the combination of descriptors of the different organs: leaves, bark, fruits and flowers. In addition to basic combination methods, we propose to consider the confusion between species, as well as predictions of affiliations in botanical taxa higher than the species.Finally, an opening chapter is devoted to the processing of these images by convolutional neural networks. Indeed, Deep Learning is increasingly used in image processing, particularly for plant organs. In this context, we propose to visualize the learned convolution filters extracting information, in order to make the link between the information extracted by these networks and botanical elements