Corpus ID: 220768638

Unsupervised Discovery of 3D Physical Objects from Video

@article{Du2020UnsupervisedDO,
  title={Unsupervised Discovery of 3D Physical Objects from Video},
  author={Yilun Du and K. Smith and Tomer Ulman and J. Tenenbaum and Jiajun Wu},
  journal={ArXiv},
  year={2020},
  volume={abs/2007.12348}
}
  • Yilun Du, K. Smith, +2 authors Jiajun Wu
  • Published 2020
  • Computer Science
  • ArXiv
  • We study the problem of unsupervised physical object discovery. Unlike existing frameworks that aim to learn to decompose scenes into 2D segments purely based on each object's appearance, we explore how physics, especially object interactions, facilitates learning to disentangle and segment instances from raw videos, and to infer the 3D geometry and position of each object, all without supervision. Drawing inspiration from developmental psychology, our Physical Object Discovery Network (POD-Net… CONTINUE READING
    2 Citations

    Figures, Tables, and Topics from this paper

    References

    SHOWING 1-10 OF 29 REFERENCES
    Unsupervised Learning of 3D Structure from Images
    • 266
    • PDF
    Multi-Object Representation Learning with Iterative Variational Inference
    • 94
    • PDF
    Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations
    • 142
    • PDF
    Learning to See Physics via Visual De-animation
    • 107
    • PDF
    Sequential Attend, Infer, Repeat: Generative Modelling of Moving Objects
    • 82
    • PDF
    Attend, Infer, Repeat: Fast Scene Understanding with Generative Models
    • 257
    • PDF
    Learning to segment moving objects in videos
    • 170
    • PDF
    Relational Neural Expectation Maximization: Unsupervised Discovery of Objects and their Interactions
    • 132
    • PDF
    Entity Abstraction in Visual Model-Based Reinforcement Learning
    • 27
    • PDF