Research map / 3D vision / Human pose estimation
COG: Confidence-aware Optimal Geometric Correspondence for Unsupervised Single-reference Novel Object Pose Estimation
Yuchen Che, Jingtu Wu, Hao Zheng, Asako Kanezaki
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 |
|---|---|---|---|
| replace argmax with confidence-marginal OT correspondence (ablation) | COG · YCB-V | mean mAP: 71.4 -> 74.3 | Improved |
| replace softmax with confidence-marginal OT correspondence (ablation) | COG · YCB-V | mean mAP: 71.5 -> 74.3 | Improved |
| use confidence marginals instead of uniform marginals in OT (ablation) | COG · YCB-V | mean mAP: 75.2 -> 75.9 | Improved |
| add semantic loss L_sem (confidence OT, no cycle loss) (ablation) | COG · YCB-V | mean mAP: 74.3 -> 75.6 | Improved |
| add cycle loss L_cycl (confidence OT, no sem loss) (ablation) | COG · YCB-V | mean mAP: 74.3 -> 74.4 | No clear effect |
| add both L_sem and L_cycl (argmax correspondence) (ablation) | COG argmax · YCB-V | mean mAP: 71.4 -> 73.1 | Improved |
| add L_sem alone with argmax correspondence (ablation) | COG argmax · YCB-V | mean mAP: 71.4 -> 70.5 | Hurt |
| add semantic priors to the affinity kernel (ablation) | COG (2 Sinkhorn iters) · YCB-V | mAP: 73.2 -> 75.9 | Improved |
| increase Sinkhorn iterations from 2 to 8 (with semantic priors) (ablation) | COG · YCB-V | mAP: 75.9 -> 75.7 | Hurt |
| add one refinement iteration at inference (ablation) | COG · BOP benchmarks | mean mAP: over 1% | Improved |
| use COG with 1% training data instead of DINO+PHS (ablation) | COG · BOP mean | mAP: 28.7 -> 64.5 | Improved |
| train on 100% instead of 5% of the data (ablation) | COG · BOP mean | mAP: 67.3 -> 68.8 | 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.
- FreeZe — FreeZe: Training-free zero-shot 6D pose estimation with geometric and vision foundation models (ECCV 2024) · better on 1 of 1 shared result
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
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.89)
- From Correspondences to Pose: Non-minimal Certifiably Optimal Relative Pose without Disambiguation (CVPR 2024 · similarity 0.87)
- Structure-Aware Correspondence Learning for Relative Pose Estimation (CVPR 2025 · similarity 0.87)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.87)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.86)
- Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence (CVPR 2024 · similarity 0.86)
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.
