Research map / 3D vision / Human pose estimation
FreeZe: Training-free zero-shot 6D pose estimation with geometric and vision foundation models
Andrea Caraffa, Davide Boscaini, Amir Hamza, Fabio Poiesi
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 |
|---|---|---|---|
| use FreeZe (training-free fusion of GeDi + DINOv2) instead of SAM6D, no refinement, CNOS masks | zero-shot 6D pose estimation · BOP core datasets mean AR | mean AR: 55.4 -> 64.0 | Improved |
| use FreeZe instead of SAM6D with refinement | zero-shot 6D pose estimation · BOP core datasets mean AR | mean AR: 63.2 -> 69.3 | Improved |
| add visual DINOv2 features to GeDi geometric features (ablation) | FreeZe · LM-O (CNOS, no ICP) | AR: 56.5 -> 64.7 | Improved |
| use GeDi instead of FCGF as geometric encoder (ablation) | FreeZe · LM-O (CNOS, no ICP) | AR: 20.8 -> 55.3 | Improved |
| use DINOv2 instead of CLIP as vision encoder (ablation) | FreeZe · LM-O (CNOS, no ICP) | AR: 30.5 -> 36.8 | Improved |
| use multi-scale GeDi (0.3, 0.4) instead of single-scale 0.3 (ablation) | FreeZe · LM-O (CNOS, no ICP) | AR: 55.3 -> 56.5 | Improved |
| scale DINOv2 backbone from ViT-L to ViT-G (ablation) | FreeZe vision encoder · LM-O (CNOS, no ICP) | AR: 36.8 -> 39.6 | Improved |
| use bounding box instead of segmentation mask localization prior, no refinement (ablation) | FreeZe · LM-O (CNOS) | AR: 64.7 -> 64.7 | No clear effect |
| use segmentation mask instead of bounding box with ICP refinement (ablation) | FreeZe · LM-O (CNOS) | AR: 68.6 -> 69.0 | Improved |
| add ICP refinement (ablation) | FreeZe · BOP core datasets mean AR | mean AR: 64.0 -> 68.8 | Improved |
| add symmetry-aware refinement (SAR) on top of ICP (ablation) | FreeZe · IC-BIN | AR: 47.3 -> 49.9 | Improved |
| add symmetry-aware refinement on datasets without symmetric textured objects (ablation) | FreeZe · LM-O, T-LESS, TUD-L, ITODD | AR: 69.0 -> 69.0 (LM-O) | No clear effect |
Later papers that compared with this method
4 later accepted papers report FreeZeV2 as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- WAPR: A Foundation Model for Wide-Angle Refinement in Unseen Object Pose Estimation (ECCV 2026 · better on 2 of 2 shared results)
- COG: Confidence-aware Optimal Geometric Correspondence for Unsupervised Single-reference Novel Object Pose Estimation (CVPR 2026 · better on 1 of 1 shared result)
- 3D-Object Perception Transformer (3PT) (CVPR 2026 · better on 0 of 1 shared result)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GigaPose — GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence (CVPR 2024) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.88)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.88)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.87)
- UA-Pose: Uncertainty-Aware 6D Object Pose Estimation and Online Object Completion with Partial References (CVPR 2025 · similarity 0.87)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.87)
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.
