Research map / 3D vision / Human pose estimation
SAC-GNC: SAmple Consensus for adaptive Graduated Non-Convexity
Valter Piedade, Chitturi Sidhartha, José Gaspar, Venu Madhav Govindu, Pedro Miraldo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive sigma search (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) | mAA (eR, 5deg): 0.465 -> 0.487 | Improved |
| add sigma initialization without sigma search (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) | mAA (eR, 5deg): 0.465 -> 0.457 | Hurt |
| add termination criteria and sigma initialization on top of sigma search (ablation) | SAC-GNC · 3DMatch HOME1 low inliers (11.6%) | Time [ms]: 7.47 -> 5.98 | Improved |
| use adaptive annealing instead of fixed gamma 5.0 (ablation) | SAC-GNC · 3DMatch HOME1 low inliers (11.6%) | mAA (eR, 5deg): 0.446 -> 0.490 | Improved |
| use adaptive annealing instead of fixed gamma 1.4 (time) (ablation) | SAC-GNC · 3DMatch HOME1 low inliers (11.6%) | Time [ms]: 7.45 -> 5.98 | Improved |
| more trials and larger queue (SAC-GNC++ vs SAC-GNC) (ablation) | SAC-GNC · 3DMatch HOME1 low inliers (11.6%) | mAA (eR, 5deg): 0.490 -> 0.505 | Improved |
| add SAC-GNC with Geman-McClure loss (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) robust loss ablation | mAA (eR, 5deg): 0.465 -> 0.487 | Improved |
| add SAC-GNC with Cauchy loss (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) robust loss ablation | mAA (eR, 5deg): 0.425 -> 0.427 | No clear effect |
| add SAC-GNC with Bisquare loss (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) robust loss ablation | mAA (eR, 5deg): 0.147 -> 0.276 | Improved |
| add SAC-GNC with Logistic loss (ablation) | vanilla GNC · 3DMatch HOME1 low inliers (11.6%) robust loss ablation | mAA (eR, 5deg): 0.162 -> 0.157 | Hurt |
| use SAC-GNC instead of GNCp | GNCp · ModelNet (50.8% inliers) | mAA (eR, 5deg): 0.627 -> 0.708 | Improved |
| use SAC-GNC instead of RANSAC | RANSAC · 3DMatch (11.6% inliers) | mAA (eR, 5deg): 0.279 -> 0.421 | Improved |
Closest papers by meaning
- Adaptive Annealing for Robust Geometric Estimation (CVPR 2023 · similarity 0.92)
- Adaptive Annealing for Robust Averaging (ECCV 2024 · similarity 0.90)
- Certifiably Optimal Outlier-Robust Geometric Perception: Semidefinite Relaxations and Scalable Global Optimization (TPAMI 2022 · similarity 0.88)
- One Ring to Rule Them All: Certifiably Robust Geometric Perception with Outliers (NeurIPS 2020 · similarity 0.88)
- Convex Relaxation for Robust Vanishing Point Estimation in Manhattan World (CVPR 2025 · similarity 0.86)
- From Variance to Veracity: Unbundling and Mitigating Gradient Variance in Differentiable Bundle Adjustment Layers (CVPR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
