Research map / 3D vision / Depth and stereo
Extreme Structure From Motion for Indoor Panoramas Without Visual Overlaps
Mohammad Amin Shabani, Weilian Song, Makoto Odamaki, Hirochika Fujiki, Yasutaka Furukawa
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 |
|---|---|---|---|
| use ground-truth room shape, room type and door/window instead of predictions (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=0.2 m | mean average precision: 30.0 -> 50.4 | Improved |
| use ground-truth room shape, room type and door/window instead of predictions (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=1.0 m | mean average precision: 67.8 -> 70.9 | Improved |
| add room-type information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=0.2 m | mean average precision: 21.3 -> 30.0 | Improved |
| add room-type information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=0.6 m | mean average precision: 60.0 -> 59.6 | No clear effect |
| add room-type information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=1.0 m | mean average precision: 64.3 -> 67.8 | Improved |
| add door/window information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=0.6 m | mean average precision: 26.5 -> 59.6 | Improved |
| add door/window information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=1.0 m | mean average precision: 34.7 -> 67.8 | Improved |
| add room-type and door/window information (ablation) | Ours · Extreme indoor SfM, mAP K=1-5, delta=0.2 m | mean average precision: 4.7 -> 30.0 | Improved |
| consider top-5 reconstructions instead of top-1 | Ours · Extreme indoor SfM, error <1.0 m | success rate (%): 47 -> 78 | Improved |
| fine-tune HorizonNet on own layout annotations | HorizonNet · Room layout on own dataset | 3D IoU: 74.42 -> 88.12 | Improved |
| use room-to-room connections to define the relational graph instead of fully connected (ablation) | ConvMPN · Extreme indoor SfM | success rate | No clear effect |
Closest papers by meaning
- MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion (CVPR 2025 · similarity 0.87)
- CasaMaestro: Multi-View Panoramas for House-Scale 3D Reconstruction (ECCV 2026 · similarity 0.87)
- Automatic 3D Indoor Scene Modeling From Single Panorama (CVPR 2018 · similarity 0.86)
- Shading Meets Motion: Self-supervised Indoor 3D Reconstruction Via Simultaneous Shape-from-Shading and Structure-from-Motion (CVPR 2025 · similarity 0.84)
- MvFFN: Multi-view Floor-Plan Feed-Forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction (NeurIPS 2026 · similarity 0.84)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · 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.
