Corpus ID: 195798944

Sim2real transfer learning for 3D pose estimation: motion to the rescue

@article{Doersch2019Sim2realTL,
  title={Sim2real transfer learning for 3D pose estimation: motion to the rescue},
  author={C. Doersch and Andrew Zisserman},
  journal={ArXiv},
  year={2019},
  volume={abs/1907.02499}
}
  • C. Doersch, Andrew Zisserman
  • Published 2019
  • Computer Science
  • ArXiv
  • Simulation is an anonymous, low-bias source of data where annotation can often be done automatically; however, for some tasks, current models trained on synthetic data generalize poorly to real data. [...] Key Result Therefore, our results suggest that motion can be a simple way to bridge a sim2real gap when video is available.Expand Abstract
    Predicting 3D Human Dynamics From Video
    20
    Controllable Attention for Structured Layered Video Decomposition
    1

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 103 REFERENCES
    MoCap-guided Data Augmentation for 3D Pose Estimation in the Wild
    177
    Towards 3D Human Pose Estimation in the Wild: A Weakly-Supervised Approach
    246
    Synthesizing Training Images for Boosting Human 3D Pose Estimation
    178
    Monocular 3D Human Pose Estimation in the Wild Using Improved CNN Supervision
    238
    Learning to Estimate 3D Human Pose and Shape from a Single Color Image
    186
    Ordinal Depth Supervision for 3D Human Pose Estimation
    115
    Structure-Aware and Temporally Coherent 3D Human Pose Estimation
    14
    A Simple Yet Effective Baseline for 3d Human Pose Estimation
    388
    Exploiting Temporal Context for 3D Human Pose Estimation in the Wild
    41