Research map / 3D vision / Depth and stereo
R-MSFM: Recurrent Multi-Scale Feature Modulation for Monocular Depth Estimating
Zhongkai Zhou, Xinnan Fan, Pengfei Shi, Yuanxue Xin
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 MSFM module at decoder beginning to the shared decoder (ablation) | R-MSFM3-A · KITTI Eigen split | Abs Rel: 0.120 -> 0.114 | Improved |
| add MSFM module after third decoder conv layer (ablation) | R-MSFM3-A · KITTI Eigen split | Abs Rel: 0.120 -> 0.118 | Improved |
| add a second MSFM module at both positions (ablation) | R-MSFM3-C · KITTI Eigen split | Abs Rel: 0.114 -> 0.115 | No clear effect |
| replace Monodepth2 U-Net decoder with parameter-shared iterative decoder and truncated encoder (ablation) | Monodepth2 · KITTI Eigen split | parameters: 14.3M -> 2.9M | Improved |
| increase iterative updates from 1 to 3 (ablation) | R-MSFM3 (1024x320) · KITTI Eigen split | Abs Rel: 0.137 -> 0.112 | Improved |
| increase iterative updates from 1 to 6 (ablation) | R-MSFM6 (1024x320) · KITTI Eigen split | Abs Rel: 0.141 -> 0.108 | Improved |
| increase updates from 5 to 6 (ablation) | R-MSFM6 (1024x320) · KITTI Eigen split | Abs Rel: 0.108 -> 0.108 | No clear effect |
| extend three updates to six updates (ablation) | R-MSFM3-C · KITTI Eigen split | Abs Rel: 0.114 -> 0.112 | Improved |
| use ImageNet pretrained encoder (ablation) | R-MSFM6 · KITTI Eigen split | Abs Rel: 0.126 -> 0.112 | Improved |
| increase input resolution from 640x192 to 1024x320 | R-MSFM6 · KITTI Eigen split | RMSE: 4.704 -> 4.470 | Improved |
| add stereo supervision (MS vs M) | R-MSFM6 · KITTI Eigen split | Abs Rel: 0.112 -> 0.111 | No clear effect |
| R-MSFM6 versus Monodepth2 in parameters and RMSE | R-MSFM6 (1024x320) · KITTI Eigen split | RMSE: 4.701 -> 4.470 | Improved |
Later papers that compared with this method
1 later accepted paper reports R-MSFM6 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- A Study on the Generality of Neural Network Structures for Monocular Depth Estimation (TPAMI 2023 · better on 2 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
- Recurrent Multiscale Feature Modulation for Geometry Consistent Depth Learning (TPAMI 2024 · similarity 0.95)
- BRNet: Exploring Comprehensive Features for Monocular Depth Estimation (ECCV 2022 · similarity 0.90)
- Self-Supervised Monocular Depth Estimation by Direction-aware Cumulative Convolution Network (ICCV 2023 · similarity 0.89)
- High-Precision Self-Supervised Monocular Depth Estimation with Rich-Resource Prior (ECCV 2024 · similarity 0.89)
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.88)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · 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.
