Research map / 3D vision / Depth and stereo
Enforcing Geometric Constraints of Virtual Normal for Depth Prediction
Wei Yin, Yifan Liu, Chunhua Shen, Youliang Yan
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 virtual normal loss (ablation) | WCEL baseline · NYUD-V2 (NYUD-Small) | rel: 0.1427 -> 0.1337 | Improved |
| add virtual normal loss (ablation) | WCEL baseline · NYUD-V2 (NYUD-Small) | rms: 0.511 -> 0.480 | Improved |
| use weighted cross-entropy instead of ordinary cross-entropy (ablation) | CEL · NYUD-V2 (NYUD-Small) | rel: 0.1456 -> 0.1427 | Improved |
| add L1 depth loss to WCEL (ablation) | WCEL · NYUD-V2 (NYUD-Small) | rel: 0.1427 -> 0.1429 | No clear effect |
| add pair-wise geometric loss (ablation) | WCEL · NYUD-V2 (NYUD-Small) | rel: 0.1427 -> 0.1380 | Improved |
| add surface normal loss (ablation) | WCEL · NYUD-V2 (NYUD-Small) | rms: 0.511 -> 0.599 | Hurt |
| add virtual normal loss on top of pair-wise loss (ablation) | WCEL+PL · NYUD-V2 (NYUD-Small) | rel: 0.1380 -> 0.1341 | Improved |
| use virtual normal loss instead of pair-wise loss (ablation) | WCEL+PL · NYUD-V2 (NYUD-Small) | rel: 0.1380 -> 0.1337 | Improved |
| increase virtual normal sample groups from 0K to 20K (ablation) | WCEL+VNL · NYUD-V2 (NYUD-Small) | rel: 5.6% | Improved |
| increase virtual normal sample groups from 20K to 100K (ablation) | WCEL+VNL · NYUD-V2 (NYUD-Small) | rel | No clear effect |
| add virtual normal loss to lightweight backbone (ablation) | MobileNetV2 model · NYUD-V2 (NYUD-Large) | rel: 0.144 -> 0.134 | Improved |
| add virtual normal loss to lightweight backbone (ablation) | MobileNetV2 model · NYUD-V2 (NYUD-Large) | delta1: 0.814 -> 0.829 | Improved |
Closest papers by meaning
- Virtual Normal: Enforcing Geometric Constraints for Accurate and Robust Depth Prediction (TPAMI 2021 · similarity 0.97)
- Adaptive Surface Normal Constraint for Geometric Estimation From Monocular Images (TPAMI 2024 · similarity 0.90)
- Adaptive Surface Normal Constraint for Depth Estimation (ICCV 2021 · similarity 0.90)
- Robust Geometry-Preserving Depth Estimation Using Differentiable Rendering (ICCV 2023 · similarity 0.89)
- CORE: Co-planarity Regularized Monocular Geometry Estimation with Weak Supervision (ICCV 2023 · similarity 0.89)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.86)
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.
