Research map / 3D vision / Depth and stereo
Auto-Rectify Network for Unsupervised Indoor Depth Estimation
Jia-Wang Bian, Huangying Zhan, Naiyan Wang, Tat-Jun Chin, Chunhua Shen, Ian D. Reid
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 rotation-removing data pre-processing (weak rectification) (ablation) | SC-Depth · NYUv2 test | AbsRel: 0.159 -> 0.143 | Improved |
| add rotation-removing data pre-processing (ablation) | Monodepth2 · NYUv2 test | AbsRel: 0.176 -> 0.151 | Improved |
| add Auto-Rectify Network on top of SC-Depth | SC-Depth · NYUv2 test | AbsRel: 0.159 -> 0.138 | Improved |
| replace data pre-processing with end-to-end ARN | SC-Depth · NYUv2 test | AbsRel: 0.143 -> 0.138 | Improved |
| add rotation-triple loss L_RT (ablation) | ARN · NYUv2 test | AbsRel: 0.150 -> 0.145 | Improved |
| add rotation-consistent loss L_RC (ablation) | ARN · NYUv2 test | AbsRel: 0.150 -> 0.148 | Improved |
| add both L_RT and L_RC (ablation) | ARN · NYUv2 test | AbsRel: 0.150 -> 0.138 | Improved |
| add ImageNet pretraining (ablation) | ARN · NYUv2 test | AbsRel: 0.164 -> 0.138 | Improved |
| add Auto-Mask to remove purely rotational pairs (ablation) | ARN · NYUv2 test | AbsRel: 0.153 -> 0.138 | Improved |
| add ARN on motion-simple driving data | SC-Depth · KITTI Eigen split | AbsRel: 0.119 -> 0.118 | No clear effect |
| add ARN trained on NYUv2, zero-shot | SC-Depth · ScanNet | AbsRel: 0.169 -> 0.156 | Improved |
| fine-tune on 7-Scenes Fire for 3 epochs | Ours · 7-Scenes Fire | AbsRel: 0.163 -> 0.105 | Improved |
Closest papers by meaning
- Moving Indoor: Unsupervised Video Depth Learning in Challenging Environments (ICCV 2019 · similarity 0.88)
- Recurrent Neural Network for (Un-)Supervised Learning of Monocular Video Visual Odometry and Depth (CVPR 2019 · similarity 0.88)
- Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue (ECCV 2016 · similarity 0.88)
- Depth Prediction without the Sensors: Leveraging Structure for Unsupervised Learning from Monocular Videos (AAAI 2019 · similarity 0.87)
- SliceNet: Deep Dense Depth Estimation From a Single Indoor Panorama Using a Slice-Based Representation (CVPR 2021 · similarity 0.87)
- Unsupervised Learning of Depth and Ego-Motion From Video (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.
