Research map / 3D vision / Human pose estimation
Pix2Pose: Pixel-Wise Coordinate Regression of Objects for 6D Pose Estimation
Kiru Park, Timothy Patten, Markus Vincze
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 transformer loss instead of plain L1 loss for symmetric objects (ablation) | Pix2Pose · T-Less obj-05 | VSD recall: 33.4 -> 55.2 | Improved |
| add transformer loss instead of L1 loss with view limits (ablation) | Pix2Pose · T-Less obj-05 | VSD recall: 47.2 -> 55.2 | Improved |
| add the second-stage refinement with re-centered bounding box (ablation) | Pix2Pose · LineMOD | ADD-10% recall: 41 -> 72.4 | Improved |
| add GAN loss to training (ablation) | Pix2Pose · LineMOD (artificially occluded test images) | fraction of correct poses | Improved |
| add GAN loss to training (ablation) | Pix2Pose · LineMOD (fully visible objects) | fraction of correct poses | No clear effect |
| use a convex hull instead of the precise 3D model for training (ablation) | Pix2Pose · LineMOD | ADD fraction | Hurt |
| use Retinanet instead of SSD-6D as 2D detector (ablation) | Pix2Pose · LineMOD | ADD-10% correct poses: 64.0 -> 71.1 | Improved |
| use Faster R-CNN instead of Retinanet as 2D detector (ablation) | Pix2Pose · LineMOD | ADD-10% correct poses: 71.1 -> 72.4 | Improved |
| use ground-truth boxes instead of Faster R-CNN detections (ablation) | Pix2Pose · LineMOD | ADD-10% correct poses: 72.4 -> 74.7 | Improved |
| detector choice, restricted to scenes with correct boxes (ablation) | Pix2Pose · LineMOD | ADD-10% correct poses: 70.9 -> 73.2 | No clear effect |
Later papers that compared with this method
8 later accepted papers report Pix2Pose as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Object-Based Visual Camera Pose Estimation From Ellipsoidal Model and 3D-Aware Ellipse Prediction (IJCV 2022 · better on 1 of 3 shared results)
- WeLSA: Learning to Predict 6D Pose from Weakly Labeled Data Using Shape Alignment (ECCV 2022 · better on 0 of 2 shared results)
- DCL-Net: Deep Correspondence Learning Network for 6D Pose Estimation (ECCV 2022 · better on 1 of 1 shared result)
- Neural Correspondence Field for Object Pose Estimation (ECCV 2022 · better on 1 of 1 shared result)
- PR-GCN: A Deep Graph Convolutional Network With Point Refinement for 6D Pose Estimation (ICCV 2021 · better on 1 of 1 shared result)
- DPODv2: Dense Correspondence-Based 6 DoF Pose Estimation (TPAMI 2021 · better on 1 of 1 shared result)
- LatentFusion: End-to-End Differentiable Reconstruction and Rendering for Unseen Object Pose Estimation (CVPR 2020 · better on 1 of 1 shared result)
- CosyPose: Consistent multi-view multi-object 6D pose estimation (ECCV 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
- SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation (ICCV 2021 · similarity 0.88)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (CVPR 2025 · similarity 0.88)
- Neural Correspondence Field for Object Pose Estimation (ECCV 2022 · similarity 0.88)
- GDRNPP: A Geometry-Guided and Fully Learning-Based Object Pose Estimator (TPAMI 2025 · similarity 0.88)
- ES6D: A Computation Efficient and Symmetry-Aware 6D Pose Regression Framework (CVPR 2022 · similarity 0.87)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.87)
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.
