Research map / 3D vision / Human pose estimation
FocalPose++: Focal Length and Object Pose Estimation via Render and Compare
Martin Cífka, Georgy Ponimatkin, Yann Labbé, Bryan Christopher Russell, Mathieu Aubry, Vladimír Petrík, Josef Šivic
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 |
|---|---|---|---|
| sample synthetic pose and focal length from a parametric distribution instead of uniform (ablation) | FocalPose++ · Pix3D sofa | MedErr_R: 2.95 -> 2.77 | Improved |
| sample from a nonparametric distribution instead of uniform (ablation) | FocalPose++ · Pix3D sofa | MedErr_R: 2.95 -> 2.82 | Improved |
| add 0.5% real images to synthetic training (ablation) | FocalPose coarse model · Pix3D sofa | MedErr_R: 5.44 -> 2.98 | Improved |
| add synthetic data to real-only training (ablation) | FocalPose coarse model · Pix3D sofa | MedErr_R: 4.13 -> 2.98 | Improved |
| raise real fraction from 0.5% to 5% (ablation) | FocalPose coarse model · Pix3D sofa | MedErr_R: 2.98 -> 3.08 | No clear effect |
| replace DINO with ML-Decoder+1k retrieval (ablation) | FocalPose++ · Pix3D model retrieval | accuracy: 62.1% -> 77.6% | Improved |
| replace DINO with ML-Decoder+1k retrieval (ablation) | FocalPose++ · Pix3D sofa | MedErr_R: 3.00 -> 2.77 | Improved |
| add 1000 synthetic images to classifier training (ablation) | ML-Decoder · Pix3D model retrieval | accuracy: 72.8% -> 77.6% | Improved |
| add 1000 synthetic images to classifier training (ablation) | ML-Decoder · Stanford Cars model retrieval | accuracy: 95.1% -> 94.7% | Hurt |
| use refined x/y update rule (ablation) | FocalPose++ · Pix3D sofa | MedErr_R: 2.81 -> 2.77 | Improved |
| add 2D reprojection loss to Huber loss (ablation) | FocalPose refiner · Pix3D sofa (FocalPose loss ablation) | MedErr_R: 6.61 -> 3.28 | Improved |
| disentangle focal length and pose in reprojection loss (ablation) | FocalPose refiner · Pix3D sofa (FocalPose loss ablation) | MedErr_R: 3.28 -> 2.98 | Improved |
Closest papers by meaning
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.87)
- SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference (NeurIPS 2025 · similarity 0.87)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · similarity 0.87)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.86)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.86)
- FAR: Flexible Accurate and Robust 6DoF Relative Camera Pose Estimation (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.
