C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion

David Novotny, Nikhila Ravi, Benjamin Graham, Natalia Neverova, Andrea Vedaldi; The IEEE International Conference on Computer Vision (ICCV), 2019, pp. 7688-7697

Abstract


We propose C3DPO, a method for extracting 3D models of deformable objects from 2D keypoint annotations in unconstrained images. We do so by learning a deep network that reconstructs a 3D object from a single view at a time, accounting for partial occlusions, and explicitly factoring the effects of viewpoint changes and object deformations. In order to achieve this factorization, we introduce a novel regularization technique. We first show that the factorization is successful if, and only if, there exists a certain canonicalization function of the reconstructed shapes. Then, we learn the canonicalization function together with the reconstruction one, which constrains the result to be consistent. We demonstrate state-of-the-art reconstruction results for methods that do not use ground-truth 3D supervision for a number of benchmarks, including Up3D and PASCAL3D+.

Related Material


[pdf] [supp]
[bibtex]
@InProceedings{Novotny_2019_ICCV,
author = {Novotny, David and Ravi, Nikhila and Graham, Benjamin and Neverova, Natalia and Vedaldi, Andrea},
title = {C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}