Research map / 3D vision / Depth and stereo
Learning Depth with Convolutional Spatial Propagation Network
Xinjing Cheng, Peng Wang, Ruigang Yang
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 |
|---|---|---|---|
| replace SPN with CSPN (no SD) (ablation) | Ma et al. baseline · NYU v2 | RMSE: 0.172 -> 0.162 | Improved |
| replace SPN with CSPN (SD preserved) (ablation) | Ma et al. baseline · NYU v2 | delta<1.02: 67.5% to 76.2% | Improved |
| add mirror connections to the encoder (ablation) | Ma et al. baseline · NYU v2 | RMSE: 0.230 -> 0.137 | Improved |
| add CSPN on top of mirror connection (ablation) | MC network · NYU v2 (SD preserved) | RMSE: 0.137 -> 0.117 | Improved |
| add SPN on top of mirror connection (ablation) | MC network · NYU v2 | RMSE: 0.137 -> 0.144 | No clear effect |
| add ASPP module (ablation) | MC+CSPN · NYU v2 | RMSE: 0.117 -> 0.116 | Improved |
| add convolutional feature fusion (CFF) (ablation) | MC+CSPN+ASPP · NYU v2 | RMSE: 0.116 -> 0.114 | Improved |
| replace ASPP with ACSPP and add CFF (ACSPF) (ablation) | MC+CSPN · NYU v2 | RMSE: 0.117 -> 0.112 | Improved |
| add bilateral filtering post-processing (ablation) | Ma et al. baseline · NYU v2 | RMSE: 0.230 -> 0.479 | Hurt |
| concatenate mirror connection after relu instead of conv or bn (ablation) | MC network · NYU v2 | RMSE: 0.531, 0.158 and 0.137 | Improved |
| add full CSPN+MC+ACSPF to baseline | Ma et al. baseline · KITTI depth completion | RMSE: 3.378 -> 2.843 | Improved |
| add 3D CSPN in disparity space instead of 2D CSPN (ablation) | PSMNet · Scene Flow | EPE: 0.992 -> 0.971 | Improved |
Closest papers by meaning
- CSPN++: Learning Context and Resource Aware Convolutional Spatial Propagation Networks for Depth Completion (AAAI 2020 · similarity 0.94)
- Depth Estimation via Affinity Learned with Convolutional Spatial Propagation Network (ECCV 2018 · similarity 0.94)
- GraphCSPN: Geometry-Aware Depth Completion via Dynamic GCNs (ECCV 2022 · similarity 0.92)
- Bilateral Propagation Network for Depth Completion (CVPR 2024 · similarity 0.91)
- Dynamic Spatial Propagation Network for Depth Completion (AAAI 2022 · similarity 0.91)
- Non-Local Spatial Propagation Network for Depth Completion (ECCV 2020 · similarity 0.90)
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.
