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

David Novotny, Nikhila Ravi, Benjamin Graham, Natalia Neverova, Andrea Vedaldi
9/5/2019
cs.CVcs.GR

Official code: facebookresearch/c3dpo_nrsfm (315 stars, MIT, last commit 2021-10-12).

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+. Source code has been made available at https://github.com/facebookresearch/c3dpo_nrsfm.

View on arXivView PDF

Code Implementations(1)

facebookresearch/c3dpo_nrsfmOfficial100%
MIT

C3DPO - Canonical 3D Pose Networks for Non-rigid Structure From Motion.

31560PythonSep 11, 20195 years agoMIT

Similar Papers

DualPM: Dual Posed-Canonical Point Maps for 3D Shape and Pose Reconstruction

86% match2024

DualPM: Dual Posed-Canonical Point Maps for 3D Shape and Pose Reconstruction

86% match2024

ShapeMatcher: Self-Supervised Joint Shape Canonicalization, Segmentation, Retrieval and Deformation

84% match1 repo232023

3DMiner: Discovering Shapes from Large-Scale Unannotated Image Datasets

84% match2023

4DRecons: 4D Neural Implicit Deformable Objects Reconstruction from a single RGB-D Camera with Geometrical and Topological Regularizations

84% match2024

Decompositional Neural Scene Reconstruction with Generative Diffusion Prior

83% match1 repo1142025

Cite this paper

@article{novotny2019cdpo,
  title  = {C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion},
  author = {David Novotny and Nikhila Ravi and Benjamin Graham and Natalia Neverova and Andrea Vedaldi},
  year   = {2019},
  eprint = {1909.02533},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1909.02533}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1909.02533.svg)](https://codeofpaper.com/paper/1909.02533)

Discussion