The Experts below are selected from a list of 3891 Experts worldwide ranked by ideXlab platform

Joseph Barfett - One of the best experts on this subject based on the ideXlab platform.

  • image augmentation using radial transform for training deep neural networks
    International Conference on Acoustics Speech and Signal Processing, 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

  • ICASSP - Image Augmentation Using Radial Transform for Training Deep Neural Networks
    2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

Hojjat Salehinejad - One of the best experts on this subject based on the ideXlab platform.

  • image augmentation using radial transform for training deep neural networks
    International Conference on Acoustics Speech and Signal Processing, 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

  • ICASSP - Image Augmentation Using Radial Transform for Training Deep Neural Networks
    2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

Shahrokh Valaee - One of the best experts on this subject based on the ideXlab platform.

  • image augmentation using radial transform for training deep neural networks
    International Conference on Acoustics Speech and Signal Processing, 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

  • ICASSP - Image Augmentation Using Radial Transform for Training Deep Neural Networks
    2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

Tim Dowdell - One of the best experts on this subject based on the ideXlab platform.

  • image augmentation using radial transform for training deep neural networks
    International Conference on Acoustics Speech and Signal Processing, 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

  • ICASSP - Image Augmentation Using Radial Transform for Training Deep Neural Networks
    2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2018
    Co-Authors: Hojjat Salehinejad, Shahrokh Valaee, Tim Dowdell, Joseph Barfett
    Abstract:

    Deep learning models have a large number of free parameters that must be estimated by efficient training of the models on a large number of training data samples to increase their generalization performance. In real-world applications, the data available to train these networks is often limited or imbalanced. We propose a sampling method based on the radial transform in a Polar Coordinate System for image augmentation to facilitate the training of deep learning models from limited source data. This pixel-wise transform provides representations of the original image in the Polar Coordinate System by generating a new image from each pixel. This technique can generate radial transformed images up to the number of pixels in the original image to increase the diversity of poorly represented image classes. Our experiments show improved generalization performance in training deep convolutional neural networks with radial transformed images.

Abbas N Moghaddam - One of the best experts on this subject based on the ideXlab platform.

  • Myocardial tagging in the Polar Coordinate System; early clinical experience.
    Journal of Cardiovascular Magnetic Resonance, 2014
    Co-Authors: Sarah N. Khan, Abbas N Moghaddam, Razieh Kaveh, Adam N. Plotnik, Evan Lehrman, Ali Nsair, J. Paul Finn
    Abstract:

    Background Quantitative MR myocardial strain analysis is typically performed using rectilinear or Cartesian grid tagging, and regional contractility is visually assessed by the deformation of the grid1. However, it is difficult to visually isolate circumferential and radial components of displacement and strain from parallel straight lines on short axis images. This study evaluates the potential of a Polar Coordinate tagging System2 for quantification of circumferential myocardial displacement in a variety of clinical conditions.

  • High resolution tagging pattern in the Polar Coordinate System; A reconstruction based approach
    2013 21st Iranian Conference on Electrical Engineering (ICEE), 2013
    Co-Authors: Ali Aghaeifar, Abbas N Moghaddam, Nafiseh Babaee, Ahmad Ayatollahi
    Abstract:

    The current gold standard for non-invasive assessment of the myocardium function is myocardium tissue tagging through cardiovascular magnetic resonance (CMR). Development of automatic tools for quantification of global and regional function of myocardium is currently an active research area and density of tag lines play a key role in this quantification. In this study we have introduced a novel framework to create high resolution tagged CMR images based on a modified reconstruction chain in MR Systems. Results show a signification potential of this framework for real-time processing of the tagged CMR images for clinical applications.

  • Polar HARP for the Polar CMR tagging
    Journal of Cardiovascular Magnetic Resonance, 2012
    Co-Authors: Nafiseh Babaee, Abbas N Moghaddam
    Abstract:

    Background The Polar Coordinate System adapts best to the morphology of the heart. The recently developed sequences that allow the CMR Tagging in the circular and radial directions facilitate the calculation and presentation of the myocardium mechanics. Development of the corresponding processing methods is required to further enhance the utilization of these sequences. Here we suggest a processing method based on the harmonic phase approach to obtain the high-resolution motion in Polar Coordinate System.

  • CMR tagging in the Polar Coordinate System
    Journal of Cardiovascular Magnetic Resonance, 2011
    Co-Authors: Abbas N Moghaddam, Yutaka Natsuaki, John Paul Finn
    Abstract:

    Strain of the myocardium is conventionally presented in the Polar Coordinate System since it adapts best to the morphology of the heart. Strain calculation would be facilitated considerably if the CMR tagging patterns were in the radial or circumferential direction. However, the CMR tagging is implemented mostly in the Cartesian Coordinate System as it is prescribed by SPAMM technique in which the gradient fields create only parallel taglines. Radial tagging is not used widely due to SAR problem for tight radial pattern. Implementation of circular tagging, to the best of our knowledge, has not been reported yet. Here we introduce an approach that makes both patterns possible for tagging based on off-resonance excitation. Its theoretical basis and practical details as well as initial results in phantoms and human hearts are presented.