Research map / 3D vision / Human pose estimation
Linear-Covariance Loss for End-to-End Learning of 6D Pose Estimation
Fulin Liu, Yinlin Hu, Mathieu Salzmann
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 LC loss to GDR-Net (ablation) | GDR-Net · LM-O | ADD(-S): 59.29 -> 66.48 | Improved |
| add LC loss to ZebraPose | ZebraPose · YCB-V | ADD(-S): 80.5 -> 82.4 | Improved |
| add LC loss to GDR-Net | GDR-Net · YCB-V | ADD(-S): 60.1 -> 70.6 | Improved |
| add the covariance term E_cov (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 45.61 -> 66.48 | Improved |
| add the prior term E_prior (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 60.69 -> 66.48 | Improved |
| keep residuals attached (not detached) in E_cov (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 61.03 -> 66.48 | Improved |
| keep weights attached (not detached) in E_cov (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 65.23 -> 66.48 | Improved |
| add the linear loss term E_linear (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 65.82 -> 66.48 | Improved |
| use 2D image-space pose representation instead of 3D (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 66.48 -> 65.99 | No clear effect |
| replace Laplace NLL with Gaussian NLL (ablation) | LC loss on GDR-Net · LM-O | ADD(-S): 66.48 -> 61.90 | Hurt |
| use quaternion, axis-angle or two-column pose representation instead of 3D (ablation) | LC loss (Laplace) · LM-O | ADD(-S): 66.48 vs 66.72, 65.85, 66.39 | No clear effect |
| add LC loss to sparse correspondence baseline with weighted PnP (ablation) | sparse MLE baseline · LM-O | ADD(-S): 59.90 -> 61.08 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- EPro-PnP — EPro-PnP: Generalized End-to-End Probabilistic Perspective-N-Points for Monocular Object Pose Estimation (CVPR 2022) · better on 3 of 3 shared results
- ZebraPose — ZebraPose: Coarse To Fine Surface Encoding for 6DoF Object Pose Estimation (CVPR 2022) · better on 3 of 3 shared results
- GDR-Net — GDR-Net: Geometry-Guided Direct Regression Network for Monocular 6D Object Pose Estimation (CVPR 2021) · better on 2 of 2 shared results
- SO-Pose — SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation (ICCV 2021) · 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
- Single-Stage 6D Object Pose Estimation (CVPR 2020 · similarity 0.88)
- Coupled Iterative Refinement for 6D Multi-Object Pose Estimation (CVPR 2022 · similarity 0.88)
- Segmentation-Driven 6D Object Pose Estimation (CVPR 2019 · similarity 0.88)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.87)
- Pos3R: 6D Pose Estimation for Unseen Objects Made Easy (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.
