Research map / Video / Video segmentation and tracking
Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 (ablation) | Basic+SSIM depth framework · KITTI raw test, 416 × 128 | AbsRel: 0.160 -> 0.158 | Helped |
| add the self-discovered consistency mask (ablation) | Basic+SSIM+GC depth framework · KITTI raw test, 416 × 128 | AbsRel: 0.158 -> 0.151 | Helped |
| add SSIM (ablation) | Basic depth framework · KITTI raw test, 416 × 128 | AbsRel: 0.161 -> 0.160 | Helped |
| use single-scale rather than 4-scale supervision (ablation) | proposed depth framework · KITTI raw test | AbsRel: 0.151 -> 0.158 | Helped |
| use 2 supervision scales instead of 1 (ablation) | proposed depth framework · KITTI raw test | AbsRel: 0.151 -> 0.152 | Hurt |
| use 3 supervision scales instead of 1 (ablation) | proposed depth framework · KITTI raw test | AbsRel: 0.151 -> 0.159 | Hurt |
| replace DispNet with DispResNet at 416 × 128 (ablation) | proposed depth framework · KITTI raw test | AbsRel: 0.151 -> 0.149 | Helped |
| increase input resolution from 416 × 128 to 832 × 256 using DispNet (ablation) | proposed depth framework · KITTI raw test | AbsRel: 0.151 -> 0.146 | Helped |
| use geometry consistency loss | depth and ego-motion networks · monocular video sequences | global scale consistency | Helped |
| use geometry inconsistency to weight photometric loss | depth framework · dynamic scenes and occlusions | depth estimation performance | Helped |
| train without geometry consistency loss (ablation) | Basic and Basic+SSIM · KITTI raw validation during training | overfitting | Hurt |
| use the proposed monocular-video method rather than CC | depth estimation · KITTI raw test (K) | AbsRel: 0.140 -> 0.137 | Helped |
Closest papers by meaning
- Unsupervised Learning of Depth and Ego-Motion From Monocular Video Using 3D Geometric Constraints (CVPR 2018 · similarity 0.91)
- Unsupervised Learning of Depth and Ego-Motion From Video (CVPR 2017 · similarity 0.89)
- Depth Prediction without the Sensors: Leveraging Structure for Unsupervised Learning from Monocular Videos (AAAI 2019 · similarity 0.89)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.89)
- Seurat: From Moving Points to Depth (CVPR 2025 · similarity 0.87)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.86)
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.