Research map / Architectures and efficiency / CNN design
Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 structured attention guided multi-scale CRF instead of the strongest baseline trained on the original set | Monocular depth estimation · NYU Depth V2 (795 training images) | rel: 0.139 -> 0.125 | Helped |
| use structured attention guided multi-scale CRF instead of Xu et al. trained with extended data | Monocular depth estimation · NYU Depth V2 (extended training data) | rel: 0.121 -> 0.125 | Hurt |
| use structured attention guided multi-scale CRF instead of the supervised monocular baseline | Monocular depth estimation · KITTI (0-80m) | rms: 4.815 -> 4.677 | Helped |
| use structured attention guided multi-scale CRF instead of the supervised monocular baseline | Monocular depth estimation · KITTI (0-80m) | delta < 1.25: 0.845 -> 0.818 | Hurt |
| add multiple deep supervision (ablation) | Front-end CNN · KITTI | rel: 0.168 -> 0.152 | Helped |
| replace naive multi-scale concatenation with CRF fusion (ablation) | Multi-scale feature fusion · KITTI | rel: 0.143 -> 0.134 | Helped |
| add unstructured attention to multi-scale CRF fusion (ablation) | Multi-scale CRF · KITTI | rel: 0.134 -> 0.127 | Helped |
| add structured constraints to attention variables (ablation) | Multi-scale CRF with attention · KITTI | rel: 0.127 -> 0.122 | Helped |
| add structured constraints to attention variables (ablation) | Multi-scale CRF with attention · KITTI | rms: 4.636 -> 4.677 | Hurt |
Closest papers by meaning
- Multi-Scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation (CVPR 2017 · similarity 0.92)
- Patch-Wise Attention Network for Monocular Depth Estimation (AAAI 2021 · similarity 0.90)
- SDC-Depth: Semantic Divide-and-Conquer Network for Monocular Depth Estimation (CVPR 2020 · similarity 0.89)
- Neural Window Fully-Connected CRFs for Monocular Depth Estimation (CVPR 2022 · similarity 0.88)
- Trap Attention: Monocular Depth Estimation With Manual Traps (CVPR 2023 · similarity 0.88)
- End-To-End Training of Hybrid CNN-CRF Models for Stereo (CVPR 2017 · similarity 0.87)
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.