Research map / 3D vision / Depth and stereo
Self-Supervised Monocular Depth Estimation: Solving the Dynamic Object Problem by Semantic Guidance
Marvin Klingner, Jan-Aike Termhlen, Jonas Mikolajczyk, Tim Fingscheidt
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 multi-task training of depth and segmentation (summed losses) (ablation) | SGDepth · KITTI split | Abs Rel: 0.108 -> 0.105 | Improved |
| add gradient scaling at the encoder (ablation) | SGDepth multi-task · KITTI split | Abs Rel: 0.105 -> 0.102 | Improved |
| add semantic masking of all DC objects (ablation) | SGDepth multi-task + scaled gradients · KITTI split | Abs Rel: 0.102 -> 0.106 | Hurt |
| add threshold to learn from non-moving DC objects (ablation) | SGDepth with semantic mask · KITTI split | Abs Rel: 0.106 -> 0.099 | Improved |
| add scheduling of fraction epsilon for unmasked images (ablation) | SGDepth add threshold · KITTI split | Abs Rel: 0.099 -> 0.097 | Improved |
| add depth training to segmentation (joint training) (ablation) | SGDepth segmentation · KITTI split | mIoU on KITTI: 43.1 -> 51.6 | Improved |
| add depth training to segmentation (joint training) (ablation) | SGDepth segmentation · Cityscapes validation | mIoU: 63.3 -> 68.2 | Improved |
| add semantic guidance (SGDepth full vs only depth) (ablation) | SGDepth · KITTI Eigen split (640x192) | Abs Rel: 0.117 -> 0.113 | Improved |
| add semantic guidance (SGDepth full vs only depth) (ablation) | SGDepth · KITTI Eigen split (1280x384) | Abs Rel: 0.113 -> 0.107 | Improved |
| add semantic guidance (SGDepth full vs only depth) (ablation) | SGDepth · KITTI Eigen split (416x128) | Abs Rel: 0.128 -> 0.121 | Improved |
| replace ResNet18 encoder with ResNet50 | SGDepth full · KITTI split | mIoU on KITTI: 50.1 -> 54.2 | Improved |
| add semantic-guided joint training (full vs only depth) on pose estimation (ablation) | SGDepth · KITTI odometry sequence 9 | ATE: 0.017 -> 0.019 | Hurt |
Closest papers by meaning
- From-Ground-To-Objects: Coarse-to-Fine Self-supervised Monocular Depth Estimation of Dynamic Objects with Ground Contact Prior (CVPR 2024 · similarity 0.91)
- S³Net: Semantic-Aware Self-supervised Depth Estimation with Monocular Videos and Synthetic Data (ECCV 2020 · similarity 0.91)
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.91)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.91)
- SC-DepthV3: Robust Self-Supervised Monocular Depth Estimation for Dynamic Scenes (TPAMI 2023 · similarity 0.91)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 2024 · similarity 0.89)
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.
