The Experts below are selected from a list of 360 Experts worldwide ranked by ideXlab platform

James C Gee - One of the best experts on this subject based on the ideXlab platform.

  • parametric medial shape representation in 3 d via the poisson partial differential equation with non linear boundary conditions
    Lecture Notes in Computer Science, 2005
    Co-Authors: Paul A Yushkevich, Hui Zhang, James C Gee
    Abstract:

    This paper presents a new shape representation for a special class of 3-D objects. In a Generative Approach to object modeling inspired by m-reps [15], skeletons of objects are explicitly defined as continuous manifolds and boundaries are derived from the skeleton by a process that involves solving a Poisson PDE with a non-linear boundary condition. This formulation helps satisfy the equality constraints that are imposed on the parameters of the representation by rules of medial geometry. One benefit of the new Approach is the ability to represent different instances of an anatomical structure using a common parametrization domain, simplifying the problem of computing correspondences between instances. Another benefit is the ability to continuously parameterize the volumetric region enclosed by the representation's boundary in a one-to-one and onto manner, in a way that preserves two of the three coordinates of the parametrization along vectors normal to the boundary. These two features make the new representation an attractive candidate for statistical analysis of shape and appearance. In this paper, the representation is carefully defined and the results of fitting the hippocampus in a deformable templates framework are presented. © Springer-Verlag Berlin Heidelberg 2005.

Paul A Yushkevich - One of the best experts on this subject based on the ideXlab platform.

  • parametric medial shape representation in 3 d via the poisson partial differential equation with non linear boundary conditions
    Lecture Notes in Computer Science, 2005
    Co-Authors: Paul A Yushkevich, Hui Zhang, James C Gee
    Abstract:

    This paper presents a new shape representation for a special class of 3-D objects. In a Generative Approach to object modeling inspired by m-reps [15], skeletons of objects are explicitly defined as continuous manifolds and boundaries are derived from the skeleton by a process that involves solving a Poisson PDE with a non-linear boundary condition. This formulation helps satisfy the equality constraints that are imposed on the parameters of the representation by rules of medial geometry. One benefit of the new Approach is the ability to represent different instances of an anatomical structure using a common parametrization domain, simplifying the problem of computing correspondences between instances. Another benefit is the ability to continuously parameterize the volumetric region enclosed by the representation's boundary in a one-to-one and onto manner, in a way that preserves two of the three coordinates of the parametrization along vectors normal to the boundary. These two features make the new representation an attractive candidate for statistical analysis of shape and appearance. In this paper, the representation is carefully defined and the results of fitting the hippocampus in a deformable templates framework are presented. © Springer-Verlag Berlin Heidelberg 2005.

Horaud Radu - One of the best experts on this subject based on the ideXlab platform.

  • Audio-Visual Speech Enhancement Using Conditional Variational Auto-Encoders
    'Institute of Electrical and Electronics Engineers (IEEE)', 2020
    Co-Authors: Sadeghi Mostafa, Leglaive Simon, Alameda-pineda Xavier, Girin Laurent, Horaud Radu
    Abstract:

    International audienceVariational auto-encoders (VAEs) are deep Generative latent variable models that can be used for learning the distribution of complex data. VAEs have been successfully used to learn a probabilistic prior over speech signals, which is then used to perform speech enhancement. One advantage of this Generative Approach is that it does not require pairs of clean and noisy speech signals at training. In this paper, we propose audio-visual variants of VAEs for single-channel and speaker-independent speech enhancement. We develop a conditional VAE (CVAE) where the audio speech Generative process is conditioned on visual information of the lip region. At test time, the audio-visual speech Generative model is combined with a noise model based on nonnegative matrix factorization, and speech enhancement relies on a Monte Carlo expectation-maximization algorithm. Experiments are conducted with the recently published NTCD-TIMIT dataset. The results confirm that the proposed audio-visual CVAE effectively fuse audio and visual information, and it improves the speech enhancement performance compared with the audio-only VAE model, especially when the speech signal is highly corrupted by noise. We also show that the proposed unsupervised audio-visual speech enhancement Approach outperforms a state-of-the-art supervised deep learning method

  • A Recurrent Variational Autoencoder for Speech Enhancement
    'Institute of Electrical and Electronics Engineers (IEEE)', 2020
    Co-Authors: Leglaive Simon, Alameda-pineda Xavier, Girin Laurent, Horaud Radu
    Abstract:

    International audienceThis paper presents a Generative Approach to speech enhancement based on a recurrent variational autoencoder (RVAE). The deep Generative speech model is trained using clean speech signals only, and it is combined with a nonnegative matrix factorization noise model for speech enhancement. We propose a variational expectation-maximization algorithm where the encoder of the RVAE is finetuned at test time, to approximate the distribution of the latent variables given the noisy speech observations. Compared with previous Approaches based on feed-forward fully-connected architectures, the proposed recurrent deep Generative speech model induces a posterior temporal dynamic over the latent variables, which is shown to improve the speech enhancement results

  • Audio-visual Speech Enhancement Using Conditional Variational Auto-Encoder
    HAL CCSD, 2020
    Co-Authors: Sadeghi Mostafa, Leglaive Simon, Alameda-pineda Xavier, Girin Laurent, Horaud Radu
    Abstract:

    Submitted to IEEE/ACM Transactions on Audio, Speech, and Language ProcessingVariational auto-encoders (VAEs) are deep Generative latent variable models that can be used for learning the distribution of complex data. VAEs have been successfully used to learn a probabilistic prior over speech signals, which is then used to perform speech enhancement. One advantage of this Generative Approach is that it does not require pairs of clean and noisy speech signals at training. In this paper, we propose audio-visual variants of VAEs for single-channel and speaker-independent speech enhancement. We develop a conditional VAE (CVAE) where the audio speech Generative process is conditioned on visual information of the lip region. At test time, the audio-visual speech Generative model is combined with a noise model based on nonnegative matrix factorization, and speech enhancement relies on a Monte Carlo expectation-maximization algorithm. Experiments are conducted with the recently published NTCD-TIMIT dataset. The results confirm that the proposed audio-visual CVAE effectively fuse audio and visual information, and it improves the speech enhancement performance compared with the audio-only VAE model, especially when the speech signal is highly corrupted by noise. We also show that the proposed unsupervised audio-visual speech enhancement Approach outperforms a state-of-the-art supervised deep learning method

Ling Shao - One of the best experts on this subject based on the ideXlab platform.

  • cycle consistent deep Generative hashing for cross modal retrieval
    IEEE Transactions on Image Processing, 2019
    Co-Authors: Yang Wang, Ling Shao
    Abstract:

    In this paper, we propose a novel deep Generative Approach to cross-modal retrieval to learn hash functions in the absence of paired training samples through the cycle consistency loss. Our proposed Approach employs adversarial training scheme to learn a couple of hash functions enabling translation between modalities while assuming the underlying semantic relationship. To induce the hash codes with semantics to the input-output pair, cycle consistency loss is further delved into the adversarial training to strengthen the correlation between the inputs and corresponding outputs. Our Approach is Generative to learn hash functions, such that the learned hash codes can maximally correlate each input–output correspondence and also regenerate the inputs so as to minimize the information loss. The learning to hash embedding is thus performed to jointly optimize the parameters of the hash functions across modalities as well as the associated Generative models. Extensive experiments on a variety of large-scale cross-modal data sets demonstrate that our proposed method outperforms the state of the arts.

  • cycle consistent deep Generative hashing for cross modal retrieval
    arXiv: Computer Vision and Pattern Recognition, 2018
    Co-Authors: Yang Wang, Ling Shao
    Abstract:

    In this paper, we propose a novel deep Generative Approach to cross-modal retrieval to learn hash functions in the absence of paired training samples through the cycle consistency loss. Our proposed Approach employs adversarial training scheme to lean a couple of hash functions enabling translation between modalities while assuming the underlying semantic relationship. To induce the hash codes with semantics to the input-output pair, cycle consistency loss is further proposed upon the adversarial training to strengthen the correlations between inputs and corresponding outputs. Our Approach is Generative to learn hash functions such that the learned hash codes can maximally correlate each input-output correspondence, meanwhile can also regenerate the inputs so as to minimize the information loss. The learning to hash embedding is thus performed to jointly optimize the parameters of the hash functions across modalities as well as the associated Generative models. Extensive experiments on a variety of large-scale cross-modal data sets demonstrate that our proposed method achieves better retrieval results than the state-of-the-arts.

Juan A Botia - One of the best experts on this subject based on the ideXlab platform.

  • a domain specific language for context modeling in context aware systems
    Journal of Systems and Software, 2013
    Co-Authors: Jose Ramon Hoyos, Jesus Garciamolina, Juan A Botia
    Abstract:

    Context-awareness refers to systems that can both sense and react based on their environment. One of the main difficulties that developers of context-aware systems must tackle is how to manage the needed context information. In this paper we present MLContext, a textual Domain-Specific Language (DSL) which is specially tailored for modeling context information. It has been implemented by applying Model-Driven Development (MDD) techniques to automatically generate software artifacts from context models. The MLContext abstract syntax has been defined as a metamodel, and model-to text transformations have been written to generate the desired software artifacts. The concrete syntax has been defined with the EMFText tool, which generates an editor and model injector. MLContext has been designed to provide a high-level abstraction, to be easy to learn, and to promote reuse of context models. A domain analysis has been applied to elicit the requirements and design choices to be taken into account in creating the DSL. As a proof of concept of the proposal, the Generative Approach has been applied to two different middleware platforms for context management.