Corpus ID: 49653092

An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution

@inproceedings{Liu2018AnIF,
  title={An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution},
  author={Rosanne Liu and Joel Lehman and Piero Molino and Felipe Petroski Such and E. Frank and A. Sergeev and J. Yosinski},
  booktitle={NeurIPS},
  year={2018}
}
  • Rosanne Liu, Joel Lehman, +4 authors J. Yosinski
  • Published in NeurIPS 2018
  • Computer Science, Mathematics
  • Few ideas have enjoyed as large an impact on deep learning as convolution. For any problem involving pixels or spatial representations, common intuition holds that convolutional neural networks may be appropriate. In this paper we show a striking counterexample to this intuition via the seemingly trivial coordinate transform problem, which simply requires learning a mapping between coordinates in (x,y) Cartesian space and one-hot pixel space. Although convolutional networks would seem… CONTINUE READING
    267 Citations
    Attention Augmented Convolutional Networks
    • 131
    • Highly Influenced
    • PDF
    A Simple Fix for Convolutional Neural Network via Coordinate Embedding
    • Highly Influenced
    • PDF
    Processing Simple Geometric Attributes with Autoencoders
    • Highly Influenced
    • PDF
    Object Parsing in Sequences Using CoordConv Gated Recurrent Networks
    • Highly Influenced
    • PDF
    Spatial Broadcast Decoder: A Simple Architecture for Learning Disentangled Representations in VAEs
    • 34
    • Highly Influenced
    • PDF
    Spatially-Adaptive Pixelwise Networks for Fast Image Translation
    • PDF
    Occlusion resistant learning of intuitive physics from videos
    • 2
    • PDF
    Efficient, Lightweight, Coordinate-Based Network for Image Super Resolution
    • 2
    Revisiting Spatial Invariance with Low-Rank Local Connectivity
    • 4
    • Highly Influenced
    • PDF

    References

    SHOWING 1-10 OF 38 REFERENCES
    Spatial Transformer Networks
    • 3,311
    • PDF
    LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
    • 622
    • Highly Influential
    • PDF
    Horovod: fast and easy distributed deep learning in TensorFlow
    • 415
    • PDF
    Measuring the Intrinsic Dimension of Objective Landscapes
    • 104
    • PDF
    Deep Image Prior
    • 716
    • PDF
    Attention is All you Need
    • 15,081
    • PDF
    Learning What and Where to Draw
    • 352
    • PDF
    Deep spatial autoencoders for visuomotor learning
    • 299
    • PDF
    Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space
    • 408
    • PDF
    DRAW: A Recurrent Neural Network For Image Generation
    • 1,391
    • Highly Influential
    • PDF