Research map / 3D vision / Human pose estimation
OrienPose: Orientation-Guided Novel View Synthesis for Single-Image Unseen Object Pose Estimation
Yating Liu, Zhaoshuai Qi, Yang Zou, Yongnan Yang, Shizhou Zhang, Yanning Zhang
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 |
|---|---|---|---|
| inject orientation priors into the NVS latent (OAG) and add orientation consistency loss | OrienPose vs NOPE · ShapeNet (10 unseen categories) | ACC 30: 52.3 -> 59.6 | Improved |
| inject orientation priors into the NVS latent (OAG) and add orientation consistency loss | OrienPose vs NOPE · ShapeNet (10 unseen categories) | Median pose error: 27.7 -> 20.4 | Improved |
| add both OAG and orientation consistency loss L_OC (vs either alone) (ablation) | OrienPose · ShapeNet bus | ACC 30: 53.3 and 55.1 (single component) -> 58.7 | Improved |
| replace NOPE's NVS network with diffusion-based Free3D | NOPE · ShapeNet (10 unseen categories) | ACC 30: 52.3 -> 40.4 | Hurt |
| zero-shot transfer from synthetic ShapeNet training to real NAVI objects | NOPE · NAVI | ACC 30: 52.3 -> 36.8 | Hurt |
| zero-shot transfer from ShapeNet to NAVI | OrienPose · ShapeNet vs NAVI | ACC 30: 59.6 -> 50.9 | Hurt |
| Gaussian blur of reference image up to 40% (ablation) | OrienPose · ShapeNet bus | ACC 30: 58.7 -> 56.0 | No clear effect |
| random occlusion of reference up to 40% of object area (ablation) | OrienPose · ShapeNet bus | ACC 30: 58.7 -> 53.3 | Hurt |
| occlusion of 10% and 20% of object area (ablation) | OrienPose · ShapeNet bus | ACC 30: 58.7 -> 57.2 / 56.9 (Table 3 occlusion columns) | No clear effect |
| provide only two views (single reference) to multi-view methods | OnePose++ and BoxDreamer · ShapeNet | ACC 30 | Hurt |
| larger viewpoint gap (45-90 deg) between reference and query | OrienPose · NAVI | qualitative pose accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- NOPE — NOPE: Novel Object Pose Estimation from a Single Image (CVPR 2024) · better on 4 of 4 shared results
- Free3D — Free3D: Consistent Novel View Synthesis without 3D Representation (CVPR 2024) · better on 3 of 3 shared results
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
- UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image (CVPR 2025 · similarity 0.90)
- Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models (ICML 2025 · similarity 0.89)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.88)
- Orient Anything V2: Unifying Orientation and Rotation Understanding (NeurIPS 2025 · similarity 0.87)
- RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects (CVPR 2025 · similarity 0.87)
- Orientation Matters: Making 3D Generative Models Orientation-Aligned (NeurIPS 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.
