Research map / 3D vision / Depth and stereo
P²Net: Patch-match and Plane-regularization for Unsupervised Indoor Depth Estimation
Zehao Yu, Lei Jin, Shenghua Gao
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 patch-match multi-view photometric consistency to the texture-weighted baseline (ablation) | P2Net · NYUv2 depth | rms: 0.786 -> 0.612 | Improved |
| add patch-match multi-view photometric consistency to the texture-weighted baseline (ablation) | P2Net · NYUv2 depth | delta<1.25: 0.628 -> 0.758 | Improved |
| add plane regularization on top of patch-match (ablation) | P2Net · NYUv2 depth | rms: 0.612 -> 0.599 | Improved |
| add plane regularization on top of patch-match (ablation) | P2Net · NYUv2 depth | delta<1.25: 0.758 -> 0.772 | Improved |
| use SURF keypoints instead of DSO (ablation) | P2Net · NYUv2 depth | rms: 0.612 -> 0.622 | No clear effect |
| use patch pattern N=3 instead of N=1 (ablation) | P2Net · NYUv2 depth | rms: 0.629 -> 0.612 | Improved |
| increase patch pattern N from 3 to 4 (ablation) | P2Net · NYUv2 depth | rms: 0.612 -> 0.634 | Hurt |
| use 5-frame input instead of 3-frame | P2Net · NYUv2 depth | rms: 0.599 -> 0.561 | Improved |
| add left-right flip test augmentation (PP) | P2Net · NYUv2 depth | rms: 0.561 -> 0.553 | Improved |
| use P2Net instead of MovingIndoor, zero-shot on ScanNet | P2Net · ScanNet depth | rms: 0.483 -> 0.420 | Improved |
| use P2Net instead of Monov2 on KITTI | P2Net · KITTI depth | rel: 0.115 vs 0.126 | Hurt |
Closest papers by meaning
- Moving Indoor: Unsupervised Video Depth Learning in Challenging Environments (ICCV 2019 · similarity 0.88)
- StructDepth: Leveraging the Structural Regularities for Self-Supervised Indoor Depth Estimation (ICCV 2021 · similarity 0.87)
- Unsupervised Learning of Geometry From Videos With Edge-Aware Depth-Normal Consistency (AAAI 2018 · similarity 0.87)
- Toward Practical Monocular Indoor Depth Estimation (CVPR 2022 · similarity 0.87)
- Geometric Structure Based and Regularized Depth Estimation From 360 Indoor Imagery (CVPR 2020 · similarity 0.86)
- Shading Meets Motion: Self-supervised Indoor 3D Reconstruction Via Simultaneous Shape-from-Shading and Structure-from-Motion (CVPR 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.
