Rotation Averaging

By Richard I. Hartley, Jochen Trumpf, Yuchao Dai, Hongdong Li. International Journal of Computer Vision, 2013

632 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/e1555d78439c96d6a386bcd1bb12fda42fd8054f

How researchers see this paper

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

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

NeRF-: Neural Radiance Fields Without Known Camera Parameters (2021, background): “In the literature, numerous approaches have been proposed to deal with this problem, for example, through rotation averaging [Bustos et al. 2019; Hartley et al. 2013].”
Canonical Capsules: Unsupervised Capsules in Canonical Pose (2020, background): “We do not compare against [41], as unfortunately no code is publicly available, and the paper is yet to be peer-reviewed.”
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age (2016, neutral): “[245], [254], structure from motion in computer vision [87], [96], [104], [168], and cryo-electron microscopy [224], [225].”
SE-Sync: A certifiably correct algorithm for synchronization over the special Euclidean group (2016, background): “Closely-related formulations also arise in structure from motion [24, 31] and camera network calibration [41] (in computer vision), sensor network localization [34], and cryo-electron microscopy [38].”
TEASER: Fast and Certifiable Point Cloud Registration (2020, background): “distance between the rotation estimate R̂ (produced by each approach) and the ground-truth R◦ [120].”
Bayesian Image Based 3D Pose Estimation (2016, background): “In particular, regarding rotation averaging, the center of mass should be within a geodesic distance no larger than π/ 2 in order to be unique, and thus well deﬁned [33–35].”
From Manifold to Manifold: Geometry-Aware Dimensionality Reduction for SPD Matrices (2014, neutral): “where L’Hôpital’s rule was used twice from (2) to (3) since the limit in (2) is indefinite.”
Sparse Inertial Poser: Automatic 3D Human Pose Estimation from Sparse IMUs (2017, background): “Actually, the squared L2-norm of e ori corresponds to the geodesic distance between ˆ R GS ( x t ) and R GS ( t ) [HTDL13,MPMR16].”
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions (2018, methodology): “We follow standard practice [29] and compute | α | as 2 cos( | α | ) = trace ( R − 1 gt R est ) − 1 , i.e ., we measure the minimum rotation angle required to align both rotations [29].”
Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems (2021, methodology): “One challenge associated with classical M-estimation [85, 86] is that the employed robust cost function ρ can be highly nonconvex, hence making local search techniques sensitive to the initial guess.”
DeepVCP: An End-to-End Deep Neural Network for Point Cloud Registration (2019, methodology): “The chordal distance [12] between R and R̄ is calculated via the Frobenius norm of the rotation matrix, denoted as ||R− R̄||F .”
Global Fusion of Relative Motions for Robust, Accurate and Scalable Structure from Motion (2013, neutral): “Alternatively, the averaging can be performed in the SO(3) Lie-group [11, 14].”
Domain Adaptation on the Statistical Manifold (2014, methodology): “Our proof then relies on two theorems from [20] stated below.”