Research map / 3D vision / Face and shape
PlaneTR: Structure-Guided Transformers for 3D Plane Recovery
Bin Tan, Nan Xue, Song Bai, Tianfu Wu, Gui-Song Xia
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 line segment tokens to the plane decoder (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 35.71 -> 38.44 | Improved |
| add line segment tokens to the plane decoder (ablation) | PlaneTR · ScanNet | VI: 0.885 -> 0.831 | Improved |
| add auxiliary plane center prediction (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 35.71 -> 38.07 | Improved |
| add auxiliary plane center prediction (ablation) | PlaneTR · ScanNet | VI: 0.885 -> 0.778 | Improved |
| add plane center prediction on top of line segments (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 38.44 -> 40.74 | Improved |
| add line segments on top of plane center prediction (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 38.07 -> 40.74 | Improved |
| increase plane queries from 20 to 30 (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 40.74 -> 40.58 | No clear effect |
| increase plane queries from 20 to 40 (ablation) | PlaneTR · ScanNet | plane recall @0.10m: 40.74 -> 39.25 | Hurt |
| use PlaneTR instead of PlaneAE | PlaneTR vs PlaneAE · ScanNet | VI: 1.025 -> 0.767 | Improved |
| use PlaneTR instead of PlaneAE | PlaneTR vs PlaneAE · NYUv2-Plane | VI: 1.393 -> 1.110 | Improved |
| use PlaneTR instead of PlaneRCNN for depth | PlaneTR vs PlaneRCNN · NYUv2 depth | Rel: 0.183 -> 0.195 | Hurt |
| use PlaneTR instead of PlaneAE for depth | PlaneTR vs PlaneAE · NYUv2 depth | delta1: 61.3 -> 63.3 | Improved |
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.
- PlaneRCNN — PlaneRCNN: 3D Plane Detection and Reconstruction From a Single Image (CVPR 2019) · better on 6 of 8 shared results
- PlaneNet — PlaneNet: Piece-Wise Planar Reconstruction From a Single RGB Image (CVPR 2018) · better on 8 of 8 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
- PlaneRecTR: Unified Query Learning for 3D Plane Recovery from a Single View (ICCV 2023 · similarity 0.88)
- Towards In-the-wild 3D Plane Reconstruction from a Single Image (CVPR 2025 · similarity 0.87)
- Recovering 3D Planes from a Single Image via Convolutional Neural Networks (ECCV 2018 · similarity 0.86)
- PlaneRAS: Learning Planar Primitives for 3D Plane Recovery (ICCV 2025 · similarity 0.86)
- Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformers (CVPR 2024 · similarity 0.85)
- AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings (CVPR 2024 · similarity 0.84)
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.
