Monocular Surface Reconstruction using 3D Deformable Part Models

Abstract : Our goal in this work is to recover an estimate of an object's surface from a single image. We address this severely ill-posed problem by employing a discriminatively-trained graphical model: we incorporate prior information about the 3D shape of an object category in terms of pairwise terms among parts, while using powerful CNN features to construct unary terms that dictate the part placement in the image. Our contributions are threefold: firstly, we extend the Deformable Part Model (DPM) paradigm to operate in a three-dimensional pose space that encodes the putative real-world coordinates of object parts. Secondly, we use branch-and-bound to perform efficient inference with DPMs, resulting in accelerations by two orders of magnitude over linear-time algorithms. Thirdly, we use Structured SVM training to properly penalize deviations between the model predictions and the 3D ground truth information during learning. Our inference requires a fraction of a second at test time and our results outperform those published recently in [17] on the PASCAL 3D+ dataset.
Liste complète des métadonnées

Cited literature [29 references]  Display  Hide  Download

https://hal.inria.fr/hal-01416479
Contributor : Stefan Kinauer <>
Submitted on : Wednesday, December 14, 2016 - 3:07:48 PM
Last modification on : Thursday, February 7, 2019 - 2:20:28 PM
Document(s) archivé(s) le : Wednesday, March 15, 2017 - 1:37:27 PM

File

0028.pdf
Files produced by the author(s)

Identifiers

Citation

Stefan Kinauer, Maxim Berman, Iasonas Kokkinos. Monocular Surface Reconstruction using 3D Deformable Part Models. "Geometry Meets Deep Learning" in ECCV 2016, Oct 2016, Amsterdam, Netherlands. pp 296-308, ⟨10.1007/978-3-319-49409-8_24⟩. ⟨hal-01416479⟩

Share

Metrics

Record views

665

Files downloads

229