Research map / 3D vision / Human pose estimation
DISP6D: Disentangled Implicit Shape and Pose Learning for Scalable 6D Pose Estimation
Yilin Wen, Xiangyu Li, Hao Pan, Lei Yang, Zheng Wang, Taku Komura, Wenping Wang
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 |
|---|---|---|---|
| condition pose codebook on shape code with tensor W instead of MLP on pose encoding only (ablation) | DISP6D · REAL275 setting III | AP30: 47.0 -> 75.3 | Improved |
| use 3rd-order tensor W instead of MLP on concatenated codes (ablation) | DISP6D · REAL275 setting III | AP30: 74.9 -> 75.3 | Improved |
| use 3rd-order tensor W instead of MLP on concatenated codes (ablation) | DISP6D · REAL275 setting III | AP10: 27.3 -> 30.9 | Improved |
| stop gradient on shape code z_o in pose code generation (ablation) | DISP6D · REAL275 setting III | AP10: 26.6 -> 30.9 | Improved |
| stop gradient on shape code z_o in pose code generation (ablation) | DISP6D · REAL275 setting III | AP60: 87.5 -> 84.3 | Hurt |
| add shape contrastive loss L_shape (ablation) | DISP6D · REAL275 setting III | AP30: 67.6 -> 75.3 | Improved |
| add shape contrastive loss L_shape (ablation) | DISP6D · REAL275 setting III | AP10: 15.2 -> 30.9 | Improved |
| condition decoder on shape code z_o instead of pose code z_p (ablation) | DISP6D · REAL275 setting III | AP30: 59.9 -> 75.3 | Improved |
| use DISP6D instead of PoseContrast for cross-category rotation | Ours-all · REAL275 (GT mask) | rotation accuracy at 30 deg: 55.0 -> 89.9 | Improved |
| use DISP6D instead of Multipath-AAE in full detection pipeline | DISP6D · T-LESS (Mask R-CNN) | VSD recall Obj 1-30: 23.51 -> 35.36 | Improved |
| use DISP6D instead of Multipath-AAE with GT mask | DISP6D · T-LESS (GT boxes) | VSD recall Obj 19-30: 59.89 -> 64.42 | Improved |
| use DISP6D instead of per-object AAE on novel objects | DISP6D · T-LESS (GT boxes) | VSD recall Obj 19-30: 66.63 -> 64.42 | Hurt |
Closest papers by meaning
- FS-Net: Fast Shape-Based Network for Category-Level 6D Object Pose Estimation With Decoupled Rotation Mechanism (CVPR 2021 · similarity 0.89)
- Disentangling Geometric Deformation Spaces in Generative Latent Shape Models (IJCV 2023 · similarity 0.89)
- CDPN: Coordinates-Based Disentangled Pose Network for Real-Time RGB-Based 6-DoF Object Pose Estimation (ICCV 2019 · similarity 0.89)
- ES6D: A Computation Efficient and Symmetry-Aware 6D Pose Regression Framework (CVPR 2022 · similarity 0.88)
- One2Any: One-Reference 6D Pose Estimation for Any Object (CVPR 2025 · similarity 0.88)
- Any6D: Model-free 6D Pose Estimation of Novel Objects (CVPR 2025 · 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.
