Research map / 3D vision / Depth and stereo
Monocular Depth Estimation Using Multi-Scale Continuous CRFs as Sequential Deep Networks
Dan Xu, Elisa Ricci, Wanli Ouyang, Xiaogang Wang, Nicu Sebe
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 multi-scale cascade CRFs on top of the front-end (ablation) | ResNet-50 · NYU Depth V2 | rel error: 0.168 -> 0.143 | Improved |
| add multi-scale cascade CRFs on top of the front-end (ablation) | VGG-CD · NYU Depth V2 | rel error: 0.203 -> 0.169 | Improved |
| use deeper front-end (ResNet-50 instead of AlexNet) (ablation) | pretrained front-end · NYU Depth V2 | rel error: 0.265 -> 0.168 | Improved |
| use symmetric encoder-decoder (VGG-CD instead of VGG-16) (ablation) | pretrained front-end · NYU Depth V2 | rel error: 0.228 -> 0.203 | Improved |
| add end-to-end CRF fusion instead of post-processing C-CRF (ablation) | VGG-CD · NYU Depth V2 | rel error: 0.193 -> 0.187 | Improved |
| add continuous CRF post-processing (ablation) | VGG-CD · NYU Depth V2 | rel error: 0.203 -> 0.193 | Improved |
| use CRF-based fusion instead of HED weighted averaging (ablation) | VGG-CD · NYU Depth V2 | rel error: 0.185 -> 0.169 | Improved |
| increase number of scales from 1 to 5 (cascade) (ablation) | cascade CRF · NYU Depth V2 | rel error: 0.187 -> 0.169 | Improved |
| use unified model instead of cascade (5-scale) (ablation) | VGG-CD · NYU Depth V2 | rel error: 0.169 -> 0.163 | Improved |
| use all-to-one message passing instead of top-down (ablation) | cascade CRF ResNet-50 · NYU Depth V2 | rel error: 0.175 -> 0.154 | Improved |
| use skip-connection instead of bottom-up message passing (ablation) | cascade CRF ResNet-50 · NYU Depth V2 | rel error: 0.169 -> 0.161 | Improved |
| add 95K image pretraining data (vs 4.7K) | ResNet-50-unified bottom up · NYU Depth V2 | rel error: 0.139 -> 0.121 | Improved |
Closest papers by meaning
- Multi-Scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation (CVPR 2017 · similarity 0.98)
- Learning Depth from Single Monocular Images Using Deep Convolutional Neural Fields (TPAMI 2016 · similarity 0.93)
- Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation (CVPR 2018 · similarity 0.93)
- Neural Window Fully-Connected CRFs for Monocular Depth Estimation (CVPR 2022 · similarity 0.89)
- Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement (ECCV 2018 · similarity 0.89)
- End-To-End Training of Hybrid CNN-CRF Models for Stereo (CVPR 2017 · 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.
