The Experts below are selected from a list of 45354 Experts worldwide ranked by ideXlab platform

Anthony Tzes - One of the best experts on this subject based on the ideXlab platform.

  • Adaptive weighted minimum prediction uncertainty control
    International Journal of Control, 1997
    Co-Authors: Anthony Tzes
    Abstract:

    The adaptive control design problem for a discrete time system with structured parameteric uncertainty, which minimizes its weighted Predicted Output uncertainty is addressed in this article. A measure of the system uncertainty is quantified through the application of the set-membership (SM) identification scheme. The SM-estimator identifies a parallelepiped within which the system parameter vector is located. For a given future input sequence, this parameter uncertainty induces an uncertainty in the Predicted system Output. The control objective is to derive the input sequence that minimizes the Predicted Output uncertainty while tracking a reference input. Simulation studies are presented to highlight the features of the proposed control scheme.

  • Weighted minimum uncertainty prediction control
    Automatica, 1996
    Co-Authors: Anthony Tzes
    Abstract:

    The control design problem for a discrete time system with structured parametric uncertainty, that minimizes its Predicted Output uncertainty is addressed in this note. The system's ARMA parameter vector coefficients are bounded within an interval. For a given future input sequence, this parameter uncertainty induces an uncertainty in the Predicted system Output. The control objective is to derive the input sequence that minimizes the Predicted Output uncertainty while tracking a reference input.

  • Adaptive weighted minimum prediction uncertainty control
    Proceedings of 35th IEEE Conference on Decision and Control, 1
    Co-Authors: Anthony Tzes
    Abstract:

    The adaptive control design problem for a discrete time system with structured parametric uncertainty, that minimizes its weighted Predicted Output uncertainty is addressed. A measure of the system uncertainty is quantified through the application of the set-membership (SM) identification scheme. The SM-estimator identifies a parallelepiped within which the system parameter vector is located. For a given future input sequence, this parameter uncertainty induces an uncertainty in the Predicted system Output. The control objective is to derive the input sequence that minimizes the Predicted Output uncertainty while tracking a reference input. Simulation studies are presented to highlight the features of the proposed control scheme.

Janis Keuper - One of the best experts on this subject based on the ideXlab platform.

  • GCPR - Object Segmentation Using Pixel-Wise Adversarial Loss
    Lecture Notes in Computer Science, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

  • Object Segmentation using Pixel-wise Adversarial Loss
    arXiv: Computer Vision and Pattern Recognition, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

Ricard Durall - One of the best experts on this subject based on the ideXlab platform.

  • GCPR - Object Segmentation Using Pixel-Wise Adversarial Loss
    Lecture Notes in Computer Science, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

  • Object Segmentation using Pixel-wise Adversarial Loss
    arXiv: Computer Vision and Pattern Recognition, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

Ullrich Köthe - One of the best experts on this subject based on the ideXlab platform.

  • GCPR - Object Segmentation Using Pixel-Wise Adversarial Loss
    Lecture Notes in Computer Science, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

  • Object Segmentation using Pixel-wise Adversarial Loss
    arXiv: Computer Vision and Pattern Recognition, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

Franz-josef Pfreundt - One of the best experts on this subject based on the ideXlab platform.

  • GCPR - Object Segmentation Using Pixel-Wise Adversarial Loss
    Lecture Notes in Computer Science, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.

  • Object Segmentation using Pixel-wise Adversarial Loss
    arXiv: Computer Vision and Pattern Recognition, 2019
    Co-Authors: Ricard Durall, Franz-josef Pfreundt, Ullrich Köthe, Janis Keuper
    Abstract:

    Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the Predicted Output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.