Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GNC-Pose: Geometry-Aware GNC-PnP for Accurate 6D Pose Estimation

Xiujin Liu

TL;DR

GNC-Pose introduces a fully learning-free monocular 6D pose estimation pipeline that leverages rendering-based initialization, geometry-aware correspondence weighting, and a GNC-driven PnP optimizer, followed by LM refinement. By voxelizing the CAD model and computing per-point geometry-consistency weights, the method robustly filters out outliers and ambiguous matches, enabling stable optimization under severe contamination. The approach achieves competitive accuracy on the YCB benchmark without any training data or category priors, outperforming traditional learning-free methods and approaching supervised methods. This framework offers a practical, interpretable alternative for robust 6D pose estimation in settings where training data is unavailable or impractical to obtain.

Abstract

We present GNC-Pose, a fully learning-free monocular 6D object pose estimation pipeline for textured objects that combines rendering-based initialization, geometry-aware correspondence weighting, and robust GNC optimization. Starting from coarse 2D-3D correspondences obtained through feature matching and rendering-based alignment, our method builds upon the Graduated Non-Convexity (GNC) principle and introduces a geometry-aware, cluster-based weighting mechanism that assigns robust per point confidence based on the 3D structural consistency of the model. This geometric prior and weighting strategy significantly stabilizes the optimization under severe outlier contamination. A final LM refinement further improve accuracy. We tested GNC-Pose on The YCB Object and Model Set, despite requiring no learned features, training data, or category-specific priors, GNC-Pose achieves competitive accuracy compared with both learning-based and learning-free methods, and offers a simple, robust, and practical solution for learning-free 6D pose estimation.

GNC-Pose: Geometry-Aware GNC-PnP for Accurate 6D Pose Estimation

TL;DR

GNC-Pose introduces a fully learning-free monocular 6D pose estimation pipeline that leverages rendering-based initialization, geometry-aware correspondence weighting, and a GNC-driven PnP optimizer, followed by LM refinement. By voxelizing the CAD model and computing per-point geometry-consistency weights, the method robustly filters out outliers and ambiguous matches, enabling stable optimization under severe contamination. The approach achieves competitive accuracy on the YCB benchmark without any training data or category priors, outperforming traditional learning-free methods and approaching supervised methods. This framework offers a practical, interpretable alternative for robust 6D pose estimation in settings where training data is unavailable or impractical to obtain.

Abstract

We present GNC-Pose, a fully learning-free monocular 6D object pose estimation pipeline for textured objects that combines rendering-based initialization, geometry-aware correspondence weighting, and robust GNC optimization. Starting from coarse 2D-3D correspondences obtained through feature matching and rendering-based alignment, our method builds upon the Graduated Non-Convexity (GNC) principle and introduces a geometry-aware, cluster-based weighting mechanism that assigns robust per point confidence based on the 3D structural consistency of the model. This geometric prior and weighting strategy significantly stabilizes the optimization under severe outlier contamination. A final LM refinement further improve accuracy. We tested GNC-Pose on The YCB Object and Model Set, despite requiring no learned features, training data, or category-specific priors, GNC-Pose achieves competitive accuracy compared with both learning-based and learning-free methods, and offers a simple, robust, and practical solution for learning-free 6D pose estimation.

Paper Structure

This paper contains 20 sections, 10 equations, 1 figure, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Learning-free 6D Pose Estimation
  4. Learning-Based 6D Pose Estimation.
  5. Robust Non-Convex Optimization and GNC.
  6. Gaps
  7. Methodology
  8. Architecture
  9. Geometry Aware Correspondence Weighting
  10. GNC-Based PnP with Geometry-Aware Inlier Selection
  11. Experiment
  12. Dataset
  13. Baseline
  14. Implementation Details
  15. Evaluation Metrics
  16. ...and 5 more sections

Figures (1)

  • Figure 1: Overall Pipeline of GNC-Pose: A fully learning-free 6D pose estimation pipeline for textured objects based on robust GNC-PnP optimization with structure-guided reweighting.