Research map / 3D vision / Depth and stereo
BRNet: Exploring Comprehensive Features for Monocular Depth Estimation
Wencheng Han, Junbo Yin, Xiaogang Jin, Xiangdong Dai, Jianbing Shen
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 detail branch (reduced stride) (ablation) | Monodepth2 · KITTI Eigen (monocular, 640x192) | Abs Rel: 0.115 -> 0.111 | Improved |
| add detail branch (reduced stride) (ablation) | Monodepth2 · KITTI Eigen (monocular, 640x192) | RMSE: 4.863 -> 4.672 | Improved |
| add global branch with element-wise summation fusion (ablation) | DB model · KITTI Eigen (monocular, 640x192) | Abs Rel: 0.111 -> 0.109 | Improved |
| replace summation with attention-based GDFM (ablation) | DB+GB model · KITTI Eigen (monocular, 640x192) | Abs Rel: 0.109 -> 0.108 | Improved |
| add residual decoder to bilateral encoder (ablation) | DB+GB model · KITTI Eigen (monocular, 640x192) | Abs Rel: 0.108 -> 0.105 | Improved |
| add residual decoder alone (ablation) | Monodepth2 · KITTI Eigen (monocular, 640x192) | RMSE: 4.863 -> 4.780 | Improved |
| feed interpolated fake-large images instead of small images | Monodepth2 · KITTI Eigen (monocular) | RMSE: 4.863 -> 4.708 | Improved |
| halve the encoder receptive field via stride | Monodepth2 · KITTI Eigen (monocular) | RMSE: 4.863 -> 4.672 | Improved |
| shrink encoder receptive field to 1/8 | Monodepth2 · KITTI Eigen (monocular) | Abs Rel: 0.111 -> 0.129 | Hurt |
| evaluate decoder outputs at different sizes (full vs 1/4) | Monodepth2 · KITTI Eigen (monocular) | Abs Rel: 0.115 vs 0.114 | No clear effect |
| use BRNet instead of Monodepth2 on far pixels | BRNet · KITTI Eigen far depth (14m+) | Abs Rel: 0.159 -> 0.147 | Improved |
| use BRNet instead of Monodepth2 on near pixels | BRNet · KITTI Eigen near depth (0-8m) | Abs Rel: 0.061 -> 0.059 | No clear effect |
Later papers that compared with this method
2 later accepted papers report BrNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · better on 1 of 1 shared result)
- Self-Supervised Monocular Depth Estimation by Direction-aware Cumulative Convolution Network (ICCV 2023 · better on 3 of 3 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
- Self-Supervised Monocular Depth Estimation by Direction-aware Cumulative Convolution Network (ICCV 2023 · similarity 0.93)
- HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation (AAAI 2021 · similarity 0.93)
- Self-Supervised Monocular Trained Depth Estimation Using Self-Attention and Discrete Disparity Volume (CVPR 2020 · similarity 0.92)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.91)
- PLADE-Net: Towards Pixel-Level Accuracy for Self-Supervised Single-View Depth Estimation With Neural Positional Encoding and Distilled Matting Loss (CVPR 2021 · similarity 0.91)
- Revealing the Reciprocal Relations Between Self-Supervised Stereo and Monocular Depth Estimation (ICCV 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.
