Recovering non-rigid 3D shape from image streams

By C. Bregler, Aaron Hertzmann, H. Biermann. 2000

920 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

3D Shape Induction from 2D Views of Multiple Objects (2016, background): “Linear shape basis (or morphable models [4, 5]) are effective for categories like faces that have a ﬁxed topology, but less so for categories with varying number of parts, e.g ., airplanes and chairs.”
Lifting Object Detection Datasets into 3D (2015, background): “These two approaches build upon traditional non-rigid structure-from-motion meth-ods [39], originally developed for reconstruction from video, and either produce sparse reconstructions [37] or have only been demonstrated on simple classes such as ﬂower petals and clown-ﬁsh, while requiring complex…”
Category-specific object reconstruction from a single image (2014, methodology): “As shown in figure 3, NRSfM allows us to reliably predict viewpoint while being robust to intraclass variations.”
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age (2016, neutral): “recover shape from non-rigid objects but with restrictive applicability, for example Pentland et al. [245, 246] require prior knowledge of some mechanical properties of the objects, Bregler et al. in [36] and [304] requires a restricted deformation of the object and show examples in human faces. Rec”
Internet Based Morphable Model (2013, methodology): “We were inspired by Bregler’s nonrigid shape factorization [14], however there it was done for a completely different problem–separating pose and shape parameters.”
Real-time 3D model acquisition (2002, background): “There has been recent work on tracking non-rigid objects [Costeira and Kanade 1998; Bregler et al. 2000], though much of it assumes either that an initial model is available or that the deformation is heavily constrained.”
Coarse-to-fine low-rank structure-from-motion (2008, methodology): “Early methods such as [5] use no prior.”
HyperNeRF (2021, neutral): “Using only a monocular video stream is convenient and inexpensive, but also introduces significant ambiguity which must be ameliorated through the use of factorization [Bregler et al. 2000] or regularization [Torresani et al. 2008].”
C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion (2019, methodology): “There are several solutions to the NR-SFM problem which can recover the viewpoint and 3D shape of a deforming object from 2D keypoints across multiple frames [4, 6, 5, 9], the majority of which are based on Bregler’s factorization framework [6].”
Simultaneous Localization, Mapping and Moving Object Tracking (2007, methodology): “The ideas based on the factorization techniques and the shape basis representation are presented in Bregler et al. (2000), Brand (2001), Torresani et al. (2001) and Xiao et al. (2004), where different constraints are used for finding the solution.”
Total Moving Face Reconstruction (2014, background): “Non-rigid structure from motion methods enable reconstruction from a single video by creating a linear basis for the non rigid motion that appears in the particular video; correspondence between the frames is typically given [14,13,20] or estimated via optical flow [23].”
Visual SLAM and Structure from Motion in Dynamic Environments (2018, methodology): “[13] proposed Nonrigid Structure from Motion (NRSfM) technique based on Tomasi-Kanade factorization under a scaled orthography camera model for the first time.”
What Shape Are Dolphins? Building 3D Morphable Models from 2D Images (2013, positive): “In a sequence of papers, Torresani and collaborators [7], [23] extend rigid-geometry SfM to nonrigid scenes using the same idea as in morphable model construction: The parameters of the model in a given image are assumed to be representable as a linear combination of the parameters of a set of 3D…”
Reconstructing PASCAL VOC (2014, background): “These two approaches build upon traditional non-rigid structure from motion methods [5], originally developed for reconstruction from video, and either produce sparse reconstructions [27] or have only been demonstrated on simple classes such as flower petals and clown-fish, while requiring complex m”
Practical low-rank matrix approximation under robust L1-norm (2012, methodology): “For example, the distribution of observed entries in a SFM problem seems highly correlated or even banded due to motion consistency, rather than randomly distributed.”
A General Framework for Motion Segmentation: Independent, Articulated, Rigid, Non-rigid, Degenerate and Non-degenerate (2006, background): “Besides rigid motions, Bregler et al.[3] and Brand[4] showed that non-rigid motions like human facial motion etc.”
PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking (2023, background): “Early motion estimation methods cast point tracking as an optimization problem defined on handcrafted features [27, 37, 57, 52, 12], and these techniques continue to drive structure-from-motion [11, 31, 43] and simultaneous localization and mapping systems [54].”