Research map / 3D vision / Depth and stereo
Can Scale-Consistent Monocular Depth Be Learned in a Self-Supervised Scale-Invariant Manner?
Lijun Wang, Yifan Wang, Linzhao Wang, Yunlong Zhan, Ying Wang, Huchuan Lu
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 mean normalization of the predicted depth (ablation) | single-stream baseline · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.118 -> 0.117 | No clear effect |
| add disentangled two-stream architecture without normalization (ablation) | baseline · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.118 -> 0.116 | No clear effect |
| add mean normalization on top of the two-stream network (ablation) | Two-stream · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.116 -> 0.112 | Improved |
| use mean instead of median normalization (ablation) | Two-stream · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.115 -> 0.112 | Improved |
| add SAG loss without scale term (ablation) | Two-stream+MN · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.112 -> 0.111 | No clear effect |
| add scale factor in SAG point cloud alignment (ablation) | SAG-w/o-scale · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.111 -> 0.109 | Improved |
| use view-synthesis correspondences instead of ICP (ablation) | SAG loss · KITTI Eigen split (per-frame scale alignment) | Abs Rel: 0.115 -> 0.109 | Improved |
| use view-synthesis correspondences instead of ICP (ablation) | SAG loss · KITTI Eigen split (per-frame scale alignment) | RMSE: 4.854 -> 4.641 | Improved |
| add SAG loss (ablation) | Two-stream+MN · KITTI Eigen split (per-sequence scale alignment) | Abs Rel: 0.124 -> 0.111 | Improved |
| add scale factor in SAG alignment (ablation) | SAG-w/o-scale · KITTI Eigen split (per-sequence scale alignment) | Abs Rel: 0.114 -> 0.111 | Improved |
| add two-stream with mean normalization (ablation) | Baseline · KITTI Eigen split (per-sequence scale alignment) | Abs Rel: 0.126 -> 0.124 | No clear effect |
| add two-stream and SAG loss (ablation) | Baseline · KITTI Eigen split (per-frame scale alignment) | RMSE: 4.956 -> 4.641 | Improved |
Closest papers by meaning
- Towards Accurate Reconstruction of 3D Scene Shape From A Single Monocular Image (TPAMI 2022 · similarity 0.89)
- Towards Zero-Shot Scale-Aware Monocular Depth Estimation (ICCV 2023 · similarity 0.89)
- Unsupervised Scale-Consistent Depth Learning from Video (IJCV 2021 · similarity 0.88)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.88)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.88)
- RSA: Resolving Scale Ambiguities in Monocular Depth Estimators through Language Descriptions (NeurIPS 2024 · similarity 0.87)
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.
