Research map / 3D vision / Human pose estimation
EPro-PnP: Generalized End-to-End Probabilistic Perspective-N-Points for Monocular Object 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 |
|---|---|---|---|
| Replace direct translation prediction with PnP-only estimation (ablation) | CDPN · LineMOD test | Mean ADD(-S): 63.21 -> 45.75 (-17.46) | Hurt |
| Add the LM solver and batch size 32 (ablation) | CDPN without translation head · LineMOD test | Mean ADD(-S): 45.75 -> 52.04 (+6.29) | Helped |
| Use KL divergence loss for end-to-end probabilistic PnP (ablation) | CDPN without translation head · LineMOD test | Mean ADD(-S): 52.04 -> 65.88 (+13.84) | Helped |
| Add derivative regularization (ablation) | Basic EPro-PnP · LineMOD test | Mean ADD(-S): 65.88 -> 67.76 (+1.88) | Helped |
| Initialize from the pretrained CDPN A1 model (ablation) | EPro-PnP with derivative regularization · LineMOD test | Mean ADD(-S): 67.76 -> 73.22 (+5.46) | Helped |
| Train for 320 additional epochs (ablation) | EPro-PnP initialized from A1 · LineMOD test | Mean ADD(-S): 73.22 -> 74.19 (+0.97) | Helped |
| Allow end-to-end coordinate learning instead of detaching coordinates (ablation) | Basic EPro-PnP · LineMOD test | Mean ADD(-S): 62.80 -> 65.88 (+3.08) | Helped |
| Allow end-to-end weight learning instead of detaching weights (ablation) | Basic EPro-PnP · LineMOD test | Mean ADD(-S): 57.19 -> 65.88 (+8.69) | Helped |
| Remove the spatial Softmax normalization denominator (ablation) | Basic EPro-PnP · LineMOD test | convergence | Hurt |
| Use Monte Carlo pose loss rather than reprojection loss (ablation) | Dense correspondence network without coordinate regression · LineMOD test | ADD(-S) 0.1d: 14.56 -> 79.46 | Helped |
| Add sparse coordinate regression supervision (ablation) | Basic EPro-PnP · nuScenes val | mAOE: 0.363 -> 0.337 | Helped |
| Remove the auxiliary reprojection loss (ablation) | Basic EPro-PnP deformable correspondence network · nuScenes val | NDS: 0.425 -> 0.408 | Hurt |
Closest papers by meaning
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.85)
- Co-op: Correspondence-based Novel Object Pose Estimation (CVPR 2025 · similarity 0.85)
- Relative Pose Estimation through Affine Corrections of Monocular Depth Priors (CVPR 2025 · similarity 0.84)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.84)
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.84)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.84)
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.