Research map / 3D vision / Human pose estimation
High-Resolution Open-Vocabulary Object 6D Pose Estimation
Jaime Corsetti, Davide Boscaini, Francesco Giuliari, Changjae Oh, Andrea Cavallaro, 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 |
|---|---|---|---|
| add crop and updated loss hyperparameters together (ablation) | Horyon · Average of 4 datasets | AR: 20.9 -> 33.4 | Improved |
| add crop (original loss hyperparameters) (ablation) | Horyon · Average of 4 datasets | AR: 20.9 -> 25.3 | Improved |
| update loss hyperparameters without crop (ablation) | Horyon · Average of 4 datasets | AR: 20.8 -> 17.8 | Hurt |
| add cross-attention fusion of visual and text (ablation) | Horyon · Average of 4 datasets | AR: 29.4 -> 33.4 | Improved |
| use cross-attention fusion instead of Oryon similarity aggregation (ablation) | Horyon · Average of 4 datasets | AR: 30.6 -> 33.4 | Improved |
| use three DINO guidance maps instead of two (ablation) | Horyon · Average of 4 datasets | AR: 31.4 -> 33.4 | Improved |
| add skip connections from DINO to decoder (ablation) | Horyon · Average of 4 datasets | AR: 30.6 -> 33.4 | Improved |
| use DINO instead of Swin-B guidance (ablation) | Horyon · Average of 4 datasets | AR: 31.3 -> 33.4 | Improved |
| use GDino VLM with token-wise features instead of CLIP (ablation) | Horyon · Average of 4 datasets | AR: 23.1 -> 31.3 | Improved |
| use GDino VLM instead of ALIGN (ablation) | Horyon · Average of 4 datasets | AR: 23.4 -> 31.3 | Improved |
| include object name in prompt (ablation) | Horyon · Average of 4 datasets | AR: 18.7 -> 33.4 | Improved |
| use correct description instead of misleading one (ablation) | Horyon · Average of 4 datasets | AR: 22.5 -> 33.4 | 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.
- LatentFusion — LatentFusion: End-to-End Differentiable Reconstruction and Rendering for Unseen Object Pose Estimation (CVPR 2020) · 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
- Open-Vocabulary Object 6D Pose Estimation (CVPR 2024 · similarity 0.93)
- FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects (CVPR 2024 · similarity 0.87)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.87)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.86)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.86)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · 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.
