Research map / 3D vision / Depth and stereo
Single-Image Piece-Wise Planar 3D Reconstruction via Associative Embedding
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use associative embedding with clustering to form plane instances rather than fixed-order plane prediction | Single-image piece-wise planar reconstruction · ScanNet test | Per-plane recall @0.05: 15.78 -> 22.93 | Helped |
| Use associative embedding with clustering to form plane instances rather than fixed-order plane prediction | Single-image piece-wise planar reconstruction · NYUv2 test | RI: 0.723 -> 0.888 | Helped |
| Use associative embedding with clustering to form plane instances rather than fixed-order plane prediction | Single-image piece-wise planar reconstruction · NYUv2 test | SC: 0.404 -> 0.519; 0.544 | Mixed |
| Add instance-level parameter supervision L_IP to pixel-level supervision (ablation) | Plane parameter estimation · ScanNet test | Per-plane recall @0.05: 20.18 -> 22.93 | Helped |
| Add pixel-level parameter supervision L_PP to instance-level supervision (ablation) | Plane parameter estimation · ScanNet test | Per-plane recall @0.05: 10.78 -> 22.93 | Helped |
| Add instance-level parameter supervision L_IP to pixel-level supervision (ablation) | Plane parameter estimation · ScanNet test | Per-pixel recall @0.05: 24.82 -> 30.59 | Helped |
| Add pixel-level parameter supervision L_PP to instance-level supervision (ablation) | Plane parameter estimation · ScanNet test | Per-pixel recall @0.05: 15.72 -> 30.59 | Helped |
| Use the proposed anchor-based mean shift instead of scikit-learn mean shift (ablation) | Plane instance clustering · ScanNet test | FPS: 2.86$^*$ -> 32.26 | Helped |
| Reduce clustering iterations from 10 to 5 with k=10 (ablation) | Plane instance clustering · ScanNet test | FPS: 32.26 -> 36.10 | Helped |
| Reduce clustering iterations from 10 to 5 with k=10 (ablation) | Plane instance clustering · ScanNet test | Per-plane recall @0.05: 22.96 -> 23.28 | No clear effect |
| Use piece-wise planar output instead of direct depth prediction (ablation) | NYUv2 depth prediction · NYUv2 test | Rel: 0.134 -> 0.141 | Hurt |
Closest papers by meaning
- AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings (CVPR 2024 · similarity 0.88)
- PlaneNet: Piece-Wise Planar Reconstruction From a Single RGB Image (CVPR 2018 · similarity 0.87)
- Towards In-the-wild 3D Plane Reconstruction from a Single Image (CVPR 2025 · similarity 0.86)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.85)
- AlphaTablets: A Generic Plane Representation for 3D Planar Reconstruction from Monocular Videos (NeurIPS 2024 · similarity 0.85)
- PLANA3R: Zero-shot Metric Planar 3D Reconstruction via Feed-forward Planar Splatting (NeurIPS 2025 · similarity 0.85)
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.