Research map / 3D vision / Depth and stereo
Hierarchical Normalization for Robust Monocular Depth Estimation
Chi Zhang, Wei Yin, Billzb Wang, Gang YU, BIN FU, Chunhua Shen
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 instance-level normalization with HDN-DR (ablation) | DPT-Hybrid depth model · 5 zero-shot benchmarks (16K train) | AbsRel: -22% mean error | Improved |
| use batch-level instead of instance-level normalization (ablation) | DPT-Hybrid depth model · 5 zero-shot benchmarks | AbsRel: 15% mean worse | Hurt |
| use only finest local normalization (Local-S) (ablation) | depth model · KITTI | AbsRel: 13.2 -> 19.3 | Hurt |
| use only finest local normalization (Local-DP) (ablation) | depth model · KITTI | AbsRel: 13.2 -> 23.3 | Hurt |
| use only local normalization (Local-DR) (ablation) | depth model · DIODE | AbsRel: 47.1 -> 25.7 | Improved |
| use HDN-S instead of instance normalization (ablation) | depth model · DIODE | AbsRel: 47.1 -> 36.4 | Improved |
| use HDN-DP instead of instance normalization (ablation) | depth model · KITTI | AbsRel: 13.2 -> 13.2 | No clear effect |
| remove random crop from training (ablation) | instance-level normalization baseline · 16K ablation | overall AbsRel | Hurt |
| use aggressive random crop (1/8, 1/4) with instance normalization (ablation) | instance-level baseline · 16K ablation | overall AbsRel | Hurt |
| add more fine-grained normalization levels (ablation) | HDN · 16K ablation | AbsRel | Mixed |
| add HDN auxiliary loss to L1 | regression model · NYU V2 fully supervised | AbsRel: 14.7 -> 13.3 | Improved |
| add SSI loss to L1 | regression model · NYU V2 fully supervised | AbsRel: 14.7 -> 14.6 | No clear effect |
Closest papers by meaning
- NDDepth: Normal-Distance Assisted Monocular Depth Estimation (ICCV 2023 · similarity 0.90)
- Virtual Normal: Enforcing Geometric Constraints for Accurate and Robust Depth Prediction (TPAMI 2021 · similarity 0.89)
- Towards Accurate Reconstruction of 3D Scene Shape From A Single Monocular Image (TPAMI 2022 · similarity 0.88)
- NDDepth: Normal-Distance Assisted Monocular Depth Estimation and Completion (TPAMI 2024 · similarity 0.88)
- DME: Unveiling the Bias for Better Generalized Monocular Depth Estimation (AAAI 2024 · similarity 0.87)
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025 · 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.
