Dynamical Pose Estimation

  title={Dynamical Pose Estimation},
  author={Heng Yang and Chris Doran and Jean-Jacques E. Slotine},
  journal={2021 IEEE/CVF International Conference on Computer Vision (ICCV)},
We study the problem of aligning two sets of 3D geometric primitives given known correspondences. Our first contribution is to show that this primitive alignment framework unifies five perception problems including point cloud registration, primitive (mesh) registration, category-level 3D registration, absolution pose estimation (APE), and category-level APE. Our second contribution is to propose DynAMical Pose estimation (DAMP), the first general and practical algorithm to solve primitive… 

Figures and Tables from this paper

Semi-automatic 3D Object Keypoint Annotation and Detection for the Masses
This work presents a full object keypoint tracking toolkit, encompassing the entire process from data collection, labeling, model learning and evaluation, and presents a semi-automatic way of collecting and labeling datasets using a wrist mounted camera on a standard robotic arm.
Certifiable Outlier-Robust Geometric Perception: Exact Semidefinite Relaxations and Scalable Global Optimization
This work proposes the first general and scalable framework to design certifiable algorithms for robust geometric perception in the presence of outliers and proposes a sparse semidefinite programming (SDP) relaxation that is much smaller than the standard Lasserre’s hierarchy while preserving exactness.


In Perfect Shape: Certifiably Optimal 3D Shape Reconstruction From 2D Landmarks
  • Heng Yang, L. Carlone
  • Computer Science, Mathematics
    2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
  • 2020
This work applies Lasserre’s hierarchy of convex Sums-of-Squares relaxations to solve the shape reconstruction problem and shows that the SOS relaxation of minimum order 2 empirically solves the original non-convex problem exactly.
Efficient Registration of High-Resolution Feature Enhanced Point Clouds
This work presents a novel framework for rigid point cloud registration, based on the principles of mechanics and thermodynamics, to precisely register high-resolution point clouds with nearly constant computational effort and without the need for pre-processing, sub-sampling or pre-alignment.
Convex Global 3D Registration with Lagrangian Duality
This paper addresses finding the globally optimal transformation in various 3D registration problems by a unified formulation that integrates common geometric registration modalities by introducing a strengthened Lagrangian dual relaxation, which surpasses previous similar approaches in effectiveness.
Optimal Pose and Shape Estimation for Category-level 3D Object Perception
The first graph-theoretic formulation to prune outliers in category-level perception, which removes outliers via convex hull and maximum clique computations is developed; the resulting approach is robust to 70 − 90% outliers.
CvxPnPL: A Unified Convex Solution to the Absolute Pose Estimation Problem from Point and Line Correspondences
A new convex method to estimate 3D pose from mixed combinations of 2D-3D point and line correspondences, the Perspective-n-Points-and-Lines problem (PnPL), is presented and the proposed relaxation allows us to recover a finite number of solutions under ambiguous configurations.
Supervised Fitting of Geometric Primitives to 3D Point Clouds
This work introduces Supervised Primitive Fitting Network (SPFN), an end-to-end neural network that can robustly detect a varying number of primitives at different scales without any user control and evaluates the approach on a novel benchmark of ANSI 3D mechanical component models.
kPAM: KeyPoint Affordances for Category-Level Robotic Manipulation
A novel formulation of category-level manipulation that uses semantic 3D keypoints as the object representation enables a simple and interpretable specification of the manipulation target as geometric costs and constraints on the keypoints, which flexibly generalizes existing pose-based manipulation methods.
Deep MANTA: A Coarse-to-Fine Many-Task Network for Joint 2D and 3D Vehicle Analysis from Monocular Image
This paper presents a novel approach, called Deep MANTA (Deep Many-Tasks), for many-task vehicle analysis from a given image, based on a new coarse-to-fine object proposal that boosts the vehicle detection.
Gravitational Approach for Point Set Registration
Experiments evidence that the new approach is robust against noise and can handle challenging scenarios with structured outliers, and is compared with the widely used Iterative Closest Point and the state of the art rigid Coherent Point Drift algorithms.
Jointly Optimizing 3D Model Fitting and Fine-Grained Classification
This work proposes to optimize 3D model fitting and fine-grained classification jointly, demonstrating the method outperforms several state-of-the-art approaches and conducting a series of analyses to explore the dependence between fine-Grained classification performance and 3D models.