Research map / 3D vision / Depth and stereo
Unsupervised Scale-Consistent Depth Learning from Video
Jia-Wang Bian, Huangying Zhan, Naiyan Wang, Zhichao Li, Le Zhang, Chunhua Shen, Ming‐Ming Cheng, Ian D. Reid
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 geometry consistency loss L_G (ablation) | baseline (ResNet18) · KITTI | AbsRel: 0.200 -> 0.155 | Improved |
| add self-discovered mask M_s on top of L_G (ablation) | SC-Depth ResNet18 · KITTI | AbsRel: 0.155 -> 0.137 | Improved |
| add auxiliary mask M_a on top of L_G (ablation) | SC-Depth ResNet18 · KITTI | AbsRel: 0.155 -> 0.153 | Improved |
| add M_a to L_G+M_s (ablation) | SC-Depth ResNet18 · KITTI | AbsRel: 0.137 -> 0.132 | Improved |
| add L_G, M_s and M_a to baseline (ablation) | SC-Depth ResNet18 · KITTI | delta1: 0.786 -> 0.835 | Improved |
| use NCC instead of SSIM (ablation) | SC-Depth ResNet18 · KITTI | AbsRel: 0.132 -> 0.145 | Hurt |
| use multi-scale supervision instead of single-scale (ablation) | SC-Depth ResNet18 · KITTI | AbsRel: 0.132 -> 0.134 (10h -> 21h training) | No clear effect |
| add geometry consistency loss L_G (ablation) | SC-Depth · KITTI Seq. 09 visual odometry | t_err (%): 12.43 -> 7.31 | Improved |
| add geometry consistency loss L_G (ablation) | Pseudo-RGBD SLAM · KITTI Seq. 09 | t_err (%): 7.81 -> 5.08 | Improved |
| add geometry consistency loss L_G (ablation) | SC-Depth · KITTI Seq. 09 depth consistency | Fitness: 0.663 -> 0.689 | Improved |
| use ResNet50 instead of ResNet18 encoder at 832x256 | SC-Depth · KITTI | AbsRel: 0.119 -> 0.114 | Improved |
| remove relative rotation (weak rectification) in training pairs | SC-Depth R50 · NYUv2 | AbsRel: 0.157 -> 0.142 | Improved |
Closest papers by meaning
- Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video (NeurIPS 2019 · similarity 0.91)
- Pseudo RGB-D for Self-Improving Monocular SLAM and Depth Prediction (ECCV 2020 · similarity 0.90)
- Depth From Videos in the Wild: Unsupervised Monocular Depth Learning From Unknown Cameras (ICCV 2019 · similarity 0.89)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.88)
- Seurat: From Moving Points to Depth (CVPR 2025 · similarity 0.88)
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025 · 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.
