Learning Task-Specific Generalized Convolutions in the Permutohedral Lattice

被引:1
|
作者
Wannenwetsch, Anne S. [1 ]
Kiefel, Martin [2 ]
Gehler, Peter V. [2 ]
Roth, Stefan [1 ]
机构
[1] Tech Univ Darmstadt, Darmstadt, Germany
[2] Amazon, Tubingen, Germany
来源
关键词
D O I
10.1007/978-3-030-33676-9_24
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Dense prediction tasks typically employ encoder-decoder architectures, but the prevalent convolutions in the decoder are not image-adaptive and can lead to boundary artifacts. Different generalized convolution operations have been introduced to counteract this. We go beyond these by leveraging guidance data to redefine their inherent notion of proximity. Our proposed network layer builds on the permutohedral lattice, which performs sparse convolutions in a high-dimensional space allowing for powerful non-local operations despite small filters. Multiple features with different characteristics span this permutohedral space. In contrast to prior work, we learn these features in a task-specific manner by generalizing the basic permutohedral operations to learnt feature representations. As the resulting objective is complex, a carefully designed framework and learning procedure are introduced, yielding rich feature embeddings in practice. We demonstrate the general applicability of our approach in different joint upsampling tasks. When adding our network layer to state-of-the-art networks for optical flow and semantic segmentation, boundary artifacts are removed and the accuracy is improved.
引用
收藏
页码:345 / 359
页数:15
相关论文
共 50 条
  • [21] Learning task-specific object recognition and scene understanding
    Drummond, T
    Caelli, T
    COMPUTER VISION AND IMAGE UNDERSTANDING, 2000, 80 (03) : 315 - 348
  • [22] Task-specific perceptual learning of texture detection and identification
    Hussain, Zahra
    Sekuler, Allison
    Bennett, Patrick
    PERCEPTION, 2010, 39 (02) : 273 - 273
  • [23] Learning Partial Power Grasp with Task-specific Contact
    Li, Miao
    2016 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND BIOMIMETICS (ROBIO), 2016, : 337 - 343
  • [24] Online Learning of Task-Specific Dynamics for Periodic Tasks
    Petric, Tadej
    Gams, Andrej
    Zlajpah, Leon
    Ude, Ales
    2014 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS 2014), 2014, : 1790 - 1795
  • [25] Task-specific perceptual learning on speed and direction discrimination
    Saffell, T
    Matthews, N
    VISION RESEARCH, 2003, 43 (12) : 1365 - 1374
  • [26] Active learning of molecular data for task-specific objectives
    Ghosh, Kunal
    Todorovic, Milica
    Vehtari, Aki
    Rinke, Patrick
    JOURNAL OF CHEMICAL PHYSICS, 2025, 162 (01):
  • [27] Learning the Curriculum with Bayesian Optimization for Task-Specific Word Representation Learning
    Tsvetkov, Yulia
    Faruqui, Manaal
    Ling, Wang
    MacWhinney, Brian
    Dyer, Chris
    PROCEEDINGS OF THE 54TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, VOL 1, 2016, : 130 - 139
  • [28] Task-specific versus generalized mnemonic representations in parietal and prefrontal cortices
    Arup Sarma
    Nicolas Y Masse
    Xiao-Jing Wang
    David J Freedman
    Nature Neuroscience, 2016, 19 : 143 - 149
  • [29] Task-specific versus generalized mnemonic representations in parietal and prefrontal cortices
    Sarma, Arup
    Masse, Nicolas Y.
    Wang, Xiao-Jing
    Freedman, David J.
    NATURE NEUROSCIENCE, 2016, 19 (01) : 143 - +
  • [30] Affordance Detection for Task-Specific Grasping Using Deep Learning
    Kokic, Mia
    Stork, Johannes A.
    Haustein, Joshua A.
    Kragic, Danica
    2017 IEEE-RAS 17TH INTERNATIONAL CONFERENCE ON HUMANOID ROBOTICS (HUMANOIDS), 2017, : 91 - 98