Corpus ID: 214667893

RAFT: Recurrent All-Pairs Field Transforms for Optical Flow

@article{Teed2020RAFTRA,
  title={RAFT: Recurrent All-Pairs Field Transforms for Optical Flow},
  author={Zachary Teed and Jun Deng},
  journal={ArXiv},
  year={2020},
  volume={abs/2003.12039}
}
  • Zachary Teed, Jun Deng
  • Published 2020
  • Computer Science
  • ArXiv
  • We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per-pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state-of-the-art performance, with strong cross-dataset generalization and high efficiency in inference time, training speed, and parameter count. Code is available… CONTINUE READING

    Figures, Tables, and Topics from this paper.

    Explore Further: Topics Discussed in This Paper

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 49 REFERENCES

    LiteFlowNet: A Lightweight Convolutional Neural Network for Optical Flow Estimation

    VIEW 7 EXCERPTS
    HIGHLY INFLUENTIAL

    SelFlow: Self-Supervised Learning of Optical Flow

    VIEW 2 EXCERPTS
    HIGHLY INFLUENTIAL

    FlowNet: Learning Optical Flow with Convolutional Networks

    VIEW 9 EXCERPTS
    HIGHLY INFLUENTIAL

    Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation

    • Junhwa Hur, Stefan Roth
    • Computer Science
    • 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
    • 2019
    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL

    Volumetric Correspondence Networks for Optical Flow

    VIEW 10 EXCERPTS
    HIGHLY INFLUENTIAL

    The Five Elements of Flow

    VIEW 2 EXCERPTS

    FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks

    VIEW 4 EXCERPTS
    HIGHLY INFLUENTIAL

    Optical Flow Estimation Using a Spatial Pyramid Network

    VIEW 2 EXCERPTS

    Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimation