Research map / 3D vision / Human pose estimation
PFRL: Pose-Free Reinforcement Learning for 6D Pose Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Refine AAE initial poses with PFRL | AAE · LINEMOD test | ADD: 31.41 -> 70.1 | Helped |
| Refine PoseCNN initial poses with PFRL | PoseCNN · LINEMOD test | ADD: 62.7 -> 79.7 | Helped |
| Refine PoseCNN initial poses with PFRL instead of DeepIM-SYN | PoseCNN · LINEMOD test | Proj. 2D: 85.3 -> 94.7 | Helped |
| Refine AAE initial poses with PFRL | AAE · T-LESS objects 19-23 | Proj. 2D: 35.12 -> 38.57 | Helped |
| Refine AAE initial poses with PFRL | AAE · T-LESS objects 19-23 | VSD: 47.47 -> 51.53 | Helped |
| Use discrete instead of continuous actions (ablation) | PFRL · LINEMOD Benchvise | Proj. 2D: 37.6 -> 58.9 (epoch 800); 60.4 -> 80.6 (epoch 3200) | Helped |
| Use discrete instead of continuous actions (ablation) | PFRL · LINEMOD Benchvise | ADD: 43.8 -> 52.8 (epoch 800); 60.9 -> 79.2 (epoch 3200) | Helped |
| Add centralization reward to IoU-difference reward (ablation) | PFRL reward · LINEMOD Iron | IoU | Helped |
| Add goal-reached reward alone to IoU-difference reward (ablation) | PFRL reward · LINEMOD Iron | IoU | Hurt |
| Combine goal-reached reward with centralization reward (ablation) | PFRL reward · LINEMOD Iron | IoU | Helped |
| Add off-policy value updates from replay to on-policy PPO (ablation) | PFRL optimization · LINEMOD Ape and Lamp | IoU | Helped |
| Train a single class-agnostic model rather than class-specific models (ablation) | PFRL · LINEMOD | ADD: 31.41 -> 48.34 mean; class-agnostic scores vary by object | Mixed |
Closest papers by meaning
- FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects (CVPR 2024 · similarity 0.86)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.86)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.86)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.85)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.85)
- 6D Object Pose Tracking in Internet Videos for Robotic Manipulation (ICLR 2025 · 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.