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
  • [31] Classification of Hyperspectral Image Based on Task-Specific Learning Network
    Ma, Xiaorui
    Ji, Sheng
    Wang, Jie
    Liu, Xiaokai
    Wang, Hongyu
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2021, 59 (10): : 8646 - 8656
  • [32] Bimodal Vein Recognition Based on Task-Specific Transfer Learning
    Wang, Guoqing
    Wang, Jun
    Pan, Zaiyu
    IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, 2017, E100D (07) : 1538 - 1541
  • [33] Evolutionary features for task-specific machine-learning applications
    Laverty, Scott
    Dey, Sourav
    Zahrt, Andrew F.
    CHEM, 2024, 10 (06): : 1623 - 1626
  • [34] TSMV: TASK-SPECIFIC MULTI-VIEW FEATURE LEARNING
    Zhang, Chengyue
    Han, Yahong
    8TH INTERNATIONAL CONFERENCE ON INTERNET MULTIMEDIA COMPUTING AND SERVICE (ICIMCS2016), 2016, : 39 - 42
  • [35] Learning task-specific discriminative representations for multiple object tracking
    Wu, Han
    Nie, Jiahao
    Zhu, Ziming
    He, Zhiwei
    Gao, Mingyu
    NEURAL COMPUTING & APPLICATIONS, 2023, 35 (10): : 7761 - 7777
  • [36] Learning Task-Specific Representation for Novel Words in Sequence Labeling
    Peng, Minlong
    Zhang, Qi
    Xing, Xiaoyu
    Gui, Tao
    Fu, Jinlan
    Huang, Xuanjing
    PROCEEDINGS OF THE TWENTY-EIGHTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2019, : 5146 - 5152
  • [37] Task-specific transfer of perceptual learning across sensory modalities
    McGovern, David P.
    Astle, Andrew T.
    Clavin, Sarah L.
    Newell, Fiona N.
    CURRENT BIOLOGY, 2016, 26 (01) : R20 - R21
  • [38] Reactive, Task-specific Object Manipulation by Metric Reinforcement Learning
    Hangl, Simon
    Ugur, Emre
    Szedmak, Sandor
    Piater, Justus
    Ude, Ales
    PROCEEDINGS OF THE 17TH INTERNATIONAL CONFERENCE ON ADVANCED ROBOTICS (ICAR), 2015, : 557 - 564
  • [39] Graph Few-shot Learning with Task-specific Structures
    Wang, Song
    Chen, Chen
    Li, Jundong
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35, NEURIPS 2022, 2022,
  • [40] TSP: Learning Task-Specific Pivots for Unsupervised Domain Adaptation
    Cui, Xia
    Coenen, Frans
    Bollegala, Danushka
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2017, PT II, 2017, 10535 : 754 - 771