Research map / 3D vision / Depth and stereo
Patch-Wise Attention Network 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 |
|---|---|---|---|
| Add patch-wise attention with patch size 64 to the baseline (ablation) | MobileNetV2 depth estimator · NYU Depth V2 | AbsRel: 0.154 -> 0.131 | Helped |
| Add patch-wise attention with patch size 64 to the baseline (ablation) | MobileNetV2 depth estimator · KITTI Eigen split | AbsRel: 0.072 -> 0.068 | Helped |
| Use patch-wise rather than global CBAM attention (ablation) | MobileNetV2 depth estimator · NYU Depth V2 | AbsRel: 0.137 -> 0.131 | Helped |
| Use patch-wise rather than global DualAttention (ablation) | MobileNetV2 depth estimator · KITTI Eigen split | AbsRel: 0.079 -> 0.068 | Helped |
| Use very small patches (sp=8) rather than patch size 64 (ablation) | Patch-wise attention · NYU Depth V2 | AbsRel: 0.131 -> 0.141 | Hurt |
| Use an image-sized patch rather than patch size 64 (ablation) | Patch-wise attention · NYU Depth V2 | AbsRel: 0.131 -> 0.145 | Hurt |
| Use patch size 128 rather than patch size 64 (ablation) | Patch-wise attention · KITTI Eigen split | AbsRel: 0.068 -> 0.070 | Hurt |
| Add DenseASPP to patch-wise attention with patch size 64 (ablation) | MobileNetV2 depth estimator · NYU Depth V2 | AbsRel: 0.131 -> 0.119 | Helped |
| Add DenseASPP to patch-wise attention with patch size 64 (ablation) | MobileNetV2 depth estimator · KITTI Eigen split | AbsRel: 0.068 -> 0.069 | No clear effect |
| Use patch-wise attention with ResNeXt101 | KITTI Eigen split, 0–50m · KITTI Eigen split, 0–50m | AbsRel: 0.058 -> 0.057 | Helped |
| Use patch-wise attention with ResNeXt101 | KITTI online benchmark · KITTI online benchmark | SILog: 11.67 -> 11.45 | Helped |
| Use patch-wise attention with DenseNet161 | NYU Depth V2 depth estimator · NYU Depth V2 | AbsRel: 0.110 -> 0.105 | Helped |
Closest papers by meaning
- Trap Attention: Monocular Depth Estimation With Manual Traps (CVPR 2023 · similarity 0.90)
- Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation (CVPR 2018 · similarity 0.90)
- Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging (CVPR 2021 · similarity 0.89)
- Depth from a Single Image by Harmonizing Overcomplete Local Network Predictions (NeurIPS 2016 · similarity 0.88)
- PatchFusion: An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation (CVPR 2024 · similarity 0.87)
- Depth Prompting for Sensor-Agnostic Depth Estimation (CVPR 2024 · 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.