Research map / 3D vision / Depth and stereo
Revealing the Reciprocal Relations Between Self-Supervised Stereo and Monocular Depth Estimation
Zhi Chen, Xiaoqing Ye, Wei Yang, Zhenbo Xu, Xiao Tan, Zhikang Zou, Errui Ding, Xinming Zhang, Liusheng Huang
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 |
|---|---|---|---|
| use Unet encoding instead of pyramid encoding in StereoNet (ablation) | StereoNet · KITTI 2015 training (stereo, after stage 1) | Abs Rel: 0.054 -> 0.052 | Improved |
| use Unet encoding instead of pyramid encoding in StereoNet (ablation) | StereoNet · KITTI 2015 training (stereo, after stage 1) | Sq Rel: 0.629 -> 0.558 | Improved |
| add occlusion-aware fusion of StereoNet and SingleNet | StereoNet · KITTI 2015 training (stereo depth) | Abs Rel: 0.052 -> 0.049 | Improved |
| distill fused disparity back into StereoNet (StereoNet-D) | StereoNet · KITTI 2015 training (stereo depth) | Abs Rel: 0.052 -> 0.048 | Improved |
| distill from StereoNet with occlusion-aware masking instead of self-supervision (ablation) | SingleNet · KITTI Eigen split (monocular depth) | Abs Rel: 0.102 -> 0.095 | Improved |
| distill from StereoNet instead of SGM proxy labels (ablation) | SingleNet · KITTI Eigen split (monocular depth) | Abs Rel: 0.100 -> 0.095 | Improved |
| set occlusion threshold alpha to 0.01 instead of 1 in distillation (ablation) | SingleNet · KITTI (SingleNet, stereo split, alpha ablation) | Abs Rel: 0.086 -> 0.083 | Improved |
| set occlusion threshold alpha to 0.01 instead of 1 in fusion (ablation) | OA Fusion · KITTI 2015 training (fusion, alpha ablation) | Abs Rel: 0.052 -> 0.049 | Improved |
| add post-processing at test time | SingleNet · KITTI Eigen split (monocular depth) | Abs Rel: 0.095 -> 0.094 | Improved |
| use StereoNet with Unet encoding instead of UnOS stereo-only | UnOS (Stereo-only) · KITTI 2015 training (stereo depth) | Abs Rel: 0.060 -> 0.052 | Improved |
Closest papers by meaning
- Reversing the cycle: self-supervised deep stereo through enhanced monocular distillation (ECCV 2020 · similarity 0.92)
- Learning Monocular Depth Estimation Infusing Traditional Stereo Knowledge (CVPR 2019 · similarity 0.92)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.92)
- Self-Supervised Monocular Trained Depth Estimation Using Self-Attention and Discrete Disparity Volume (CVPR 2020 · similarity 0.91)
- Self-Supervised Deep Monocular Depth Estimation With Ambiguity Boosting (TPAMI 2021 · similarity 0.91)
- Self-Supervised Learning of Depth Inference for Multi-View Stereo (CVPR 2021 · similarity 0.91)
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.
