Single Image 3D Without a Single 3D Image

Abstract

Do we really need 3D labels in order to learn how to predict 3D? In this paper, we show that one can learn a mapping from appearance to 3D properties without ever seeing a single explicit 3D label. Rather than use explicit supervision, we use the regularity of indoor scenes to learn the mapping in a completely unsupervised manner. We demonstrate this on both a standard 3D scene understanding dataset as well as Internet images for which 3D is unavailable, precluding supervised learning. Despite never seeing a 3D label, our method produces competitive results.

Cite

Text

Fouhey et al. "Single Image 3D Without a Single 3D Image." International Conference on Computer Vision, 2015. doi:10.1109/ICCV.2015.126

Markdown

[Fouhey et al. "Single Image 3D Without a Single 3D Image." International Conference on Computer Vision, 2015.](https://mlanthology.org/iccv/2015/fouhey2015iccv-single/) doi:10.1109/ICCV.2015.126

BibTeX

@inproceedings{fouhey2015iccv-single,
  title     = {{Single Image 3D Without a Single 3D Image}},
  author    = {Fouhey, David F. and Hussain, Wajahat and Gupta, Abhinav and Hebert, Martial},
  booktitle = {International Conference on Computer Vision},
  year      = {2015},
  doi       = {10.1109/ICCV.2015.126},
  url       = {https://mlanthology.org/iccv/2015/fouhey2015iccv-single/}
}