Research map / 3D vision / Gaussian splatting
PoseGaussian: 6D Pose Estimation for Unseen Objects via Sparse-View Object-Level 3D Gaussian Splatting
Wubin Shi, Shaoyan Gai, Feipeng Da
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 density control (ablation) | PoseGaussian · Linemod | Pose ACC: 85.0 -> 87.5 | Improved |
| add view-warping augmentation (ablation) | PoseGaussian · Linemod | Pose ACC: 85.9 -> 87.5 | Improved |
| use improved 3DGS (depth prior init and sparse training) instead of vanilla 3DGS (ablation) | PoseGaussian · Linemod | Pose ACC: 82.1 -> 87.5 | Improved |
| use two-stage learning-guided ICP initializer instead of Gen6D selector (ablation) | PoseGaussian · Linemod | Pose ACC: 68.7 -> 87.5 | Improved |
| use two-stage learning-guided ICP initializer instead of Cas6D selector (ablation) | PoseGaussian · Linemod | Pose ACC: 71.4 -> 87.5 | Improved |
| use 3DGS differentiable rendering refiner instead of learning-based Gen6D refiner (ablation) | PoseGaussian · Linemod | Pose ACC: 79.2 -> 87.5 | Improved |
| reduce reference views from over 200 to 8 (ablation) | PoseGaussian · Linemod ADD(S)-0.1d mean | ADD(S)-0.1d mean: 87.2 -> 54.30 | Hurt |
| use 8 reference views instead of 16 for baselines (comparison with Gen6D/Cas6D/OnePose++ at 16 views) | PoseGaussian · Linemod (sparse views) | ADD(S)-0.1d mean: 54.30 vs 47.83 (Cas6D, 16 views) | Improved |
| use 8 reference views while baselines use 16 views | PoseGaussian · GenMOP | ADD-0.1d average: 27.56 vs 37.01 (Cas6D, 16 views) | Hurt |
| add depth structural prior and sparse training strategy to 3DGS under few views (ablation) | 3DGS reconstruction · Linemod, 8-32 views | pose accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- IG-6DoF — iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting (CVPR 2025) · better on 3 of 3 shared results
- OnePose++ — OnePose: One-Shot Object Pose Estimation Without CAD Models (CVPR 2022) · better on 1 of 1 shared result
- Gen6D — Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images (ECCV 2022) · 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
- iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting (CVPR 2025 · similarity 0.93)
- No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images (ICLR 2025 · similarity 0.90)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.89)
- SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- KeyGS: A Keyframe-Centric Gaussian Splatting Method for Monocular Image Sequences (AAAI 2025 · similarity 0.89)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.89)
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.
