Research map / 3D vision / Depth and stereo
DP-SfM: Dual-Pixel Structure-From-Motion Without Scale Ambiguity
Lilika Makabe, Kaneyoshi Ashida, Hiroaki Santo, Fumio Okura, Yasuyuki Matsushita
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 the intensity-based optimization stage after the linear initialization (ablation) | DP-SfM · Scale estimation, 6 scenes | average scale error e_s: 0.266 -> 0.153 | Improved |
| use multi-aperture images pooled per focal length | DP-SfM · Scale estimation | average scale error: 15.5% -> 9.8% | Improved |
| add intensity-based optimization in multi-aperture setting (ablation) | DP-SfM · Scale estimation, multi-aperture | e_s: 0.223 -> 0.098 | Improved |
| add intensity-based optimization at 50mm in multi-aperture setting (ablation) | DP-SfM · Scene2 and Scene3, 50mm | scale ratio: 1.232 -> 1.322 (Scene2); 1.351 -> 1.515 (Scene3) | Hurt |
| use multi-aperture inputs for Scene4 DSLR 50mm | DP-SfM · Scene4 | scale ratio: 1.630 -> 1.046 | Improved |
| use DP-SfM instead of UniDepth v2 metric depth | UniDepth v2 · Scale estimation, six scenes | e_s: 0.416 -> 0.153 | Improved |
| use DP-SfM instead of DepthAnything v2 metric depth | DepthAnything v2 · Scale estimation, six scenes | e_s: 0.519 -> 0.153 | Improved |
| use DP-SfM instead of ZoeDepth | ZoeDepth · Scale estimation, six scenes | e_s: 0.565 -> 0.153 | Improved |
| change lens (35mm, 50mm, 85mm) | DP-SfM · Scale estimation, DSLR | e_s | No clear effect |
| reduce aperture size to f/8.0 on 50mm lens (ablation) | DP-SfM · Scale estimation, DSLR 50mm | e_s: 0.127 -> 0.383 | Hurt |
| use smartphone (small fixed aperture f/1.73) instead of DSLR | DP-SfM · Scale estimation | e_s: 0.284 vs 0.153 overall | Hurt |
Closest papers by meaning
- Resolving Scale Ambiguity in Multi-view 3D Reconstruction using Dual-Pixel Sensors (ECCV 2024 · similarity 0.96)
- Multi-view Reconstruction via SfM-guided Monocular Depth Estimation (CVPR 2025 · similarity 0.85)
- DVP-MVS: Synergize Depth-Edge and Visibility Prior for Multi-View Stereo (AAAI 2025 · similarity 0.85)
- Dual-Level Precision Edges Guided Multi-View Stereo with Accurate Planarization (AAAI 2025 · similarity 0.85)
- Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion (CVPR 2024 · similarity 0.85)
- Passive Snapshot Coded Aperture Dual-Pixel RGB-D Imaging (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.
