Research map / 3D vision / Depth and stereo
Unsupervised High-Resolution Depth Learning From Videos With Dual Networks
Junsheng Zhou, Yuwang Wang, Kaihuai Qin, Wenjun Zeng
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 HR-Net and SA module on top of LR-Net (ablation) | dual network depth model · KITTI Eigen split | Abs Rel: 0.132 -> 0.121 | Improved |
| add HR-Net to LR-Net (ablation) | dual network depth model · KITTI Eigen split | Abs Rel: 0.132 -> 0.123 | Improved |
| add SA-Attention module between LR-Net and HR-Net (ablation) | dual network depth model · KITTI Eigen split | Sq Rel: 0.881 -> 0.837 | Improved |
| add HR-Net (ablation) | dual network depth model · KITTI pixels >20m depth | Abs Rel: 0.199 -> 0.187 | Improved |
| add HR-Net (ablation) | dual network depth model · KITTI pixels <=20m depth | Abs Rel: 0.108 -> 0.102 | Improved |
| add HR-Net (ablation) | dual network depth model · KITTI Eigen test split (depth sharpness) | Tenengrad score: 20784.4 -> 106987.2 | Improved |
| add SA module (ablation) | dual network depth model · KITTI road region (low texture) | Abs Rel: 0.085 -> 0.080 | Improved |
| add SA module (ablation) | dual network depth model · KITTI tree and grass region (textured) | Abs Rel: 0.190 -> 0.192 | No clear effect |
| use ImageNet pretraining (ablation) | dual network depth model · KITTI Eigen split | Abs Rel: 0.135 -> 0.121 | Improved |
| use dual ResNet50 network instead of ResNet18 encoder at 384x1248 (ablation) | depth model · KITTI Eigen split | Abs Rel: 0.138 -> 0.121 | Improved |
| use dual ResNet50 network instead of MobileNet V2 encoder (ablation) | depth model · KITTI Eigen split | Abs Rel: 0.147 -> 0.121 | Improved |
| add HR-Net to the depth network used with PoseNet (ablation) | dual network depth model · KITTI Odometry sequence 09 | ATE (m): 0.017 -> 0.015 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GeoNet — GeoNet: Geometric Neural Network for Joint Depth and Surface Normal Estimation (CVPR 2018) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation (AAAI 2021 · similarity 0.90)
- BRNet: Exploring Comprehensive Features for Monocular Depth Estimation (ECCV 2022 · similarity 0.90)
- Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging (CVPR 2021 · similarity 0.89)
- Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue (ECCV 2016 · similarity 0.89)
- Refine and Distill: Exploiting Cycle-Inconsistency and Knowledge Distillation for Unsupervised Monocular Depth Estimation (CVPR 2019 · similarity 0.88)
- RA-Depth: Resolution Adaptive Self-Supervised Monocular Depth Estimation (ECCV 2022 · 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.
