Research map / 3D vision / Human pose estimation
MixRI: Mixing Features of Reference Images for Novel Object Pose Estimation
Xinhang Liu, Jiawei Shi, Zheng Dang, Yuchao Dai
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 Self-Attention between Points (SAP) (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L mean (attention ablation) | mean AR: 2.6 -> 11.9 | Improved |
| add Mix-Attention between Reference & Query (MARQ) (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L mean (attention ablation) | mean AR: 2.6 -> 22.8 | Improved |
| add both SAP and MARQ (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L mean (attention ablation) | mean AR: 2.6 -> 39.5 | Improved |
| add MARQ on top of SAP (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L mean (attention ablation) | mean AR: 11.9 -> 39.5 | Improved |
| enable occlusion flags with tau_occ 0.8 (ablation) | MixRI · BOP 7 core datasets (no refinement), occlusion threshold | mean AR: 30.7 -> 34.1 | Improved |
| use tau_occ 0.8 instead of 0.5 (ablation) | MixRI · BOP 7 core datasets (no refinement), occlusion threshold | mean AR: 31.8 -> 34.1 | Improved |
| increase reference images from 12 to 24 at test time | MixRI · BOP 7 core datasets (no refinement) | mean AR: 31.4 -> 34.1 | Improved |
| increase test-time reference images beyond 60 (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L | AR | Hurt |
| increase number of correspondences up to about 2400 (ablation) | MixRI · BOP YCB-V/LM-O/TUD-L | AR | Improved |
| use MixRI instead of GigaPose with 4 reference images | GigaPose · BOP YCB-V/LM-O/TUD-L mean | mean AR: 9.1 -> 17.9 | Improved |
| use MixRI instead of GigaPose with 24 reference images | GigaPose · BOP YCB-V/LM-O/TUD-L mean | mean AR: 18.3 -> 39.5 | Improved |
| add MegaPose refinement to MixRI (12 ref) | MixRI · BOP 7 core datasets (no refinement) | mean AR: 31.4 -> 45.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GigaPose — GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence (CVPR 2024) · better on 7 of 9 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
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.87)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.87)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.86)
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.85)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.85)
- MatchU: Matching Unseen Objects for 6D Pose Estimation from RGB-D Images (CVPR 2024 · similarity 0.85)
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.
