Índice
Literatura académica sobre el tema "Apprentissage profond géométrique"
Crea una cita precisa en los estilos APA, MLA, Chicago, Harvard y otros
Consulte las listas temáticas de artículos, libros, tesis, actas de conferencias y otras fuentes académicas sobre el tema "Apprentissage profond géométrique".
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.
Tesis sobre el tema "Apprentissage profond géométrique"
Mazari, Ahmed. "Apprentissage profond pour la reconnaissance d’actions en vidéos". Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS171.
Texto completoNowadays, video contents are ubiquitous through the popular use of internet and smartphones, as well as social media. Many daily life applications such as video surveillance and video captioning, as well as scene understanding require sophisticated technologies to process video data. It becomes of crucial importance to develop automatic means to analyze and to interpret the large amount of available video data. In this thesis, we are interested in video action recognition, i.e. the problem of assigning action categories to sequences of videos. This can be seen as a key ingredient to build the next generation of vision systems. It is tackled with AI frameworks, mainly with ML and Deep ConvNets. Current ConvNets are increasingly deeper, data-hungrier and this makes their success tributary of the abundance of labeled training data. ConvNets also rely on (max or average) pooling which reduces dimensionality of output layers (and hence attenuates their sensitivity to the availability of labeled data); however, this process may dilute the information of upstream convolutional layers and thereby affect the discrimination power of the trained video representations, especially when the learned action categories are fine-grained
Maignant, Elodie. "Plongements barycentriques pour l'apprentissage géométrique de variétés : application aux formes et graphes". Electronic Thesis or Diss., Université Côte d'Azur, 2023. http://www.theses.fr/2023COAZ4096.
Texto completoAn MRI image has over 60,000 pixels. The largest known human protein consists of around 30,000 amino acids. We call such data high-dimensional. In practice, most high-dimensional data is high-dimensional only artificially. For example, of all the images that could be randomly generated by coloring 256 x 256 pixels, only a very small subset would resemble an MRI image of a human brain. This is known as the intrinsic dimension of such data. Therefore, learning high-dimensional data is often synonymous with dimensionality reduction. There are numerous methods for reducing the dimension of a dataset, the most recent of which can be classified according to two approaches.A first approach known as manifold learning or non-linear dimensionality reduction is based on the observation that some of the physical laws behind the data we observe are non-linear. In this case, trying to explain the intrinsic dimension of a dataset with a linear model is sometimes unrealistic. Instead, manifold learning methods assume a locally linear model.Moreover, with the emergence of statistical shape analysis, there has been a growing awareness that many types of data are naturally invariant to certain symmetries (rotations, reparametrizations, permutations...). Such properties are directly mirrored in the intrinsic dimension of such data. These invariances cannot be faithfully transcribed by Euclidean geometry. There is therefore a growing interest in modeling such data using finer structures such as Riemannian manifolds. A second recent approach to dimension reduction consists then in generalizing existing methods to non-Euclidean data. This is known as geometric learning.In order to combine both geometric learning and manifold learning, we investigated the method called locally linear embedding, which has the specificity of being based on the notion of barycenter, a notion a priori defined in Euclidean spaces but which generalizes to Riemannian manifolds. In fact, the method called barycentric subspace analysis, which is one of those generalizing principal component analysis to Riemannian manifolds, is based on this notion as well. Here we rephrase both methods under the new notion of barycentric embeddings. Essentially, barycentric embeddings inherit the structure of most linear and non-linear dimension reduction methods, but rely on a (locally) barycentric -- affine -- model rather than a linear one.The core of our work lies in the analysis of these methods, both on a theoretical and practical level. In particular, we address the application of barycentric embeddings to two important examples in geometric learning: shapes and graphs. In addition to practical implementation issues, each of these examples raises its own theoretical questions, mostly related to the geometry of quotient spaces. In particular, we highlight that compared to standard dimension reduction methods in graph analysis, barycentric embeddings stand out for their better interpretability. In parallel with these examples, we characterize the geometry of locally barycentric embeddings, which generalize the projection computed by locally linear embedding. Finally, algorithms for geometric manifold learning, novel in their approach, complete this work
Girard, Nicolas. "Approches d'apprentissage et géométrique pour l'extraction automatique d'objets à partir d'images de télédétection". Thesis, Université Côte d'Azur, 2020. https://tel.archives-ouvertes.fr/tel-03177997.
Texto completoCreating a digital double of the Earth in the form of maps has many applications in e.g. autonomous driving, automated drone delivery, urban planning, telecommunications, and disaster management. Geographic Information Systems (GIS) are the frameworks used to integrate geolocalized data and represent maps. They represent shapes of objects in a vector representation so that it is as sparse as possible while representing shapes accurately, as well as making it easier to edit than raster data. With the increasing amount of satellite and aerial images being captured every day, automatic methods are being developed to transfer the information found in those remote sensing images into Geographic Information Systems. Deep learning methods for image segmentation are able to delineate the shapes of objects found in images however they do so with a raster representation, in the form of a mask. Post-processing vectorization methods then convert that raster representation into a vector representation compatible with GIS. Another challenge in remote sensing is to deal with a certain type of noise in the data, which is the misalignment between different layers of geolocalized information (e.g. between images and building cadaster data). This type of noise is frequent due to various errors introduced during the processing of remote sensing data. This thesis develops combined learning and geometric approaches with the purpose to improve automatic GIS mapping from remote sensing images.We first propose a method for correcting misaligned maps over images, with the first motivation for them to match, but also with the motivation to create remote sensing datasets for image segmentation with alignment-corrected ground truth. Indeed training a model on misaligned ground truth would not lead to great performance, whereas aligned ground truth annotations will result in better models. During this work we also observed a denoising effect of our alignment model and use it to denoise a misaligned dataset in a self-supervised manner, meaning only the misaligned dataset was used for training.We then propose a simple approach to use a neural network to directly output shape information in the vector representation, in order to by-pass the post-processing vectorization step. Experimental results on a dataset of solar panels show that the proposed network succeeds in learning to regress polygon coordinates, yielding directly vectorial map outputs. Our simple method is limited to predicting polygons with a fixed number of vertices though.While more recent methods for learning directly in the vector representation do not have this limitation, they still have other limitations in terms of the type of object shapes they can predict. More complex topological cases such as objects with holes or buildings touching each other (with a common wall which is very typical of European city centers) are not handled by these fully deep learning methods. We thus propose a hybrid approach alleviating those limitations by training a neural network to output a segmentation probability map as usual and also to output a frame field aligned with the contours of detected objects (buildings in our case). That frame field constitutes additional shape information learned by the network. We then propose our highly parallelizable polygonization method for leveraging that frame field information to vectorize the segmentation probability map efficiently. Because our polygonization method has access to additional information in the form of a frame field, it can be less complex than other advanced vectorization methods and is thus faster. Lastly, requiring an image segmentation network to also output a frame field only adds two convolutional layers and virtually does not increase inference time, making the use of a frame field only beneficial
Fang, Hao. "Modélisation géométrique à différent niveau de détails d'objets fabriqués par l'homme". Thesis, Université Côte d'Azur (ComUE), 2019. http://www.theses.fr/2019AZUR4002/document.
Texto completoGeometric modeling of man-made objects from 3D data is one of the biggest challenges in Computer Vision and Computer Graphics. The long term goal is to generate a CAD-style model in an as-automatic-as-possible way. To achieve this goal, difficult issues have to be addressed including (i) the scalability of the modeling process with respect to massive input data, (ii) the robustness of the methodology to various defect-laden input measurements, and (iii) the geometric quality of output models. Existing methods work well to recover the surface of free-form objects. However, in case of manmade objects, it is difficult to produce results that approach the quality of high-structured representations as CAD models.In this thesis, we present a series of contributions to the field. First, we propose a classification method based on deep learning to distinguish objects from raw 3D point cloud. Second, we propose an algorithm to detect planar primitives in 3D data at different level of abstraction. Finally, we propose a mechanism to assemble planar primitives into compact polygonal meshes. These contributions are complementary and can be used sequentially to reconstruct city models at various level-of-details from airborne 3D data. We illustrate the robustness, scalability and efficiency of our methods on both laser and multi-view stereo data composed of man-made objects
Mehr, Éloi. "Unsupervised Learning of 3D Shape Spaces for 3D Modeling". Electronic Thesis or Diss., Sorbonne université, 2019. http://www.theses.fr/2019SORUS566.
Texto completoEven though 3D data is becoming increasingly more popular, especially with the democratization of virtual and augmented experiences, it remains very difficult to manipulate a 3D shape, even for designers or experts. Given a database containing 3D instances of one or several categories of objects, we want to learn the manifold of plausible shapes in order to develop new intelligent 3D modeling and editing tools. However, this manifold is often much more complex compared to the 2D domain. Indeed, 3D surfaces can be represented using various embeddings, and may also exhibit different alignments and topologies. In this thesis we study the manifold of plausible shapes in the light of the aforementioned challenges, by deepening three different points of view. First of all, we consider the manifold as a quotient space, in order to learn the shapes’ intrinsic geometry from a dataset where the 3D models are not co-aligned. Then, we assume that the manifold is disconnected, which leads to a new deep learning model that is able to automatically cluster and learn the shapes according to their typology. Finally, we study the conversion of an unstructured 3D input to an exact geometry, represented as a structured tree of continuous solid primitives
Hosni, Nadia. "De l’analyse en composantes principales fonctionnelle à l’autoencodeur convolutif profond sur les trajectoires de formes de Kendall pour l’analyse et la reconnaissance de la démarche en 3D". Thesis, Lille 1, 2020. http://www.theses.fr/2020LIL1I066.
Texto completoIn the field of Computer Vision and Pattern Recognition, human behavior understanding has attracted the attention of several research groups and specialized companies. Successful intelligent solutions will be playing an important role in applications which involve humanrobot or human-computer interaction, biometrics recognition (security), and physical performance assessment (healthcare and well-being) since it will help the human beings were their cognitive and limited capabilities cannot perform well. In my thesis project, we investigate the problem of 3D gait recognition and analysis as gait is user-friendly and a well-accepted technology especially with the availability of RGB-D sensors and algorithms for detecting and tracking of human landmarks in video streams. Unlike other biometrics such as fingerprints, face or iris, it can be acquired at a large distance and do not require any collaboration of the end user. This point makes gait recognition suitable in intelligent video surveillance problems used, for example, in the security field as one of the behavioral biometrics or in healthcare as good physical patterns. However, using 3D human body tracked landmarks to provide such motions’ analysis faces many challenges like spatial and temporal variations and high dimension. Hence, in this thesis, we propose novel frameworks to infer 3D skeletal sequences for the purpose of 3D gait analysis and recognition. They are based on viewing the above-cited sequences as time-parameterized trajectories on the Kendall shape space S, results of modding out shape-preserving transformations, i.e., scaling, translation and rotation. Considering the non-linear structure of the manifold on which these shape trajectories are lying, the use of the conventional machine learning tools and the standard computational tools cannot be straightforward. Hence, we make use of geometric steps related to the Riemannian geometry in order to handle the problem of nonlinearity. Our first contribution is a geometric-functional framework for 3D gait analysis with a direct application to behavioral biometric recognition and physical performance assessment. We opt for an extension of the functional Principal Component Analysis to the underlying space. This functional analysis of trajectories, grounding on the geometry of the space of representation, allows to extract compact and efficient biometric signatures. In addition, we also propose a geometric deep convolutional auto-encoder (DCAE) for the purpose of gait recognition from time-varying 3D skeletal data. To accommodate the Neural Network architectures to obtained manifold-valued trajectories on the underlying non-linear space S, these trajectories are mapped to a certain vector space by means of someRiemannien geometry tools, prior to the encoding-decoding scheme. Without applying any prior temporal alignment step (e.g., Dynamic Time Warping) or modeling (e.g., HMM, RNN), they are then fed to a convolutional auto-encoder to build an identity-relevant latent space that showed discriminating capacities for identifying persons when no Temporal Alignment is applied to the time-parametrized gait trajectories: Efficient gait patterns are extracted. Both approaches were tested on several publicly available datasets and shows promising results
Poulenard, Adrien. "Structures for deep learning and topology optimization of functions on 3D shapes". Thesis, Institut polytechnique de Paris, 2020. http://www.theses.fr/2020IPPAX007.
Texto completoThe field of geometry processing is following a similar path as image analysis with the explosion of publications dedicated to deep learning in recent years. An important research effort is being made to reproduce the successes of deep learning 2D computer vision in the context of 3D shape analysis. Unlike images shapes comes in various representations like meshes or point clouds which often lack canonical structure. This makes traditional deep learning algorithms like Convolutional Neural Networks (CNN) non straightforward to apply to 3D data. In this thesis we propose three main contributions:First, we introduce a method to compare functions on different domains without correspondences and to deform them to make the topology of their set of levels more alike. We apply our method to the classical problem of shape matching in the context of functional maps to produce smoother and more accurate correspondences. Furthermore, our method is based on the continuous optimization of a differentiable energy with respect to the compared functions and is applicable to deep learning. We make two direct contributions to deep learning on 3D data. We introduce a new convolution operator over triangles meshes based on local polar coordinates and apply it to deep learning on meshes. Unlike previous works our operator takes all choices of polar coordinates into account without loss of directional information. Lastly we introduce a new rotation invariant convolution layer over point clouds and show that CNNs based on this layer can outperform state of the art methods in standard tasks on un-alligned datasets even with data augmentation
Madra, Anna. "Analyse et visualisation de la géométrie des matériaux composites à partir de données d’imagerie 3D". Thesis, Compiègne, 2017. http://www.theses.fr/2017COMP2387/document.
Texto completoThe subject of the thesis project between Laboratoire Roberval at Université de Technologie Compiègne and Center for High-Performance Composites at Ecole Polytechnique de Montréal considered the design of a deep learning architecture with semantics for automatic generation of models of composite materials microstructure based on X-ray microtomographic imagery. The thesis consists of three major parts. Firstly, the methods of microtomographic image processing are presented, with an emphasis on phase segmentation. Then, the geometric features of phase elements are extracted and used to classify and identify new morphologies. The method is presented for composites filled with short natural fibers. The classification approach is also demonstrated for the study of defects in composites, but with spatial features added to the process. A high-level descriptor "defect genome" is proposed, that permits comparison of the state o defects between specimens. The second part of the thesis introduces structural segmentation on the example of woven reinforcement in a composite. The method relies on dual kriging, calibrated by the segmentation error from learning algorithms. In the final part, a stochastic formulation of the kriging model is presented based on Gaussian Processes, and distribution of physical properties of a composite microstructure is retrieved, ready for numerical simulation of the manufacturing process or of mechanical behavior