Research map / Label-efficient and robust learning / Domain adaptation and generalization
DESC: Domain Adaptation for Depth Estimation via Semantic Consistency
Adrian Lopez-Rodriguez, Krystian Mikolajczyk
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 image translation to source-only (ablation) | depth model · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.223 -> 0.199 | Improved |
| add image translation (ablation) | depth model · Virtual KITTI->KITTI (Eigen, cap 80m) | Sq Rel: 2.205 -> 2.436 | Hurt |
| add instance-height pseudo-labels to Img. (ablation) | depth model · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.199 -> 0.171 | Improved |
| add semantic+edge consistency constraint to Img. (ablation) | depth model · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.199 -> 0.173 | Improved |
| feed semantic map in addition to edges to G_S (ablation) | DESC · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.187 -> 0.173 | Improved |
| combine pseudo-labels and consistency (full DESC) (ablation) | DESC · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.173 -> 0.156 | Improved |
| increase smoothing weight to 0.1 without consistency (ablation) | +Img.+Ins. · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.171 -> 0.165 | Improved |
| use per-instance learned heights instead of one height per class (ablation) | DESC · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.160 -> 0.156 | Improved |
| use known target focal length vs unknown f_T (ablation) | DESC · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.156 -> 0.156 | No clear effect |
| learn scalar phi rather than fix phi=1 (ablation) | DESC · Cityscapes->KITTI (Eigen, cap 80m) | Abs Rel: 0.169 -> 0.149 | Improved |
| add stereo supervision in KITTI | DESC · Virtual KITTI->KITTI (Eigen, cap 80m) | Abs Rel: 0.156 -> 0.119 | Improved |
| use DESC instead of T2Net | T2Net · Virtual KITTI->KITTI (Eigen, cap 80m) | Sq Rel: 1.396 -> 1.067 | Improved |
Closest papers by meaning
- 3D-PL: Domain Adaptive Depth Estimation with 3D-Aware Pseudo-Labeling (ECCV 2022 · similarity 0.91)
- Domain Adaptive Semantic Segmentation With Self-Supervised Depth Estimation (ICCV 2021 · similarity 0.91)
- Geometry-Aware Symmetric Domain Adaptation for Monocular Depth Estimation (CVPR 2019 · similarity 0.90)
- Mind The Edge: Refining Depth Edges in Sparsely-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.90)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.88)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.88)
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.
