Research map / 3D vision / Depth and stereo
GeoNet++: Iterative Geometric Neural Network with Edge-Aware Refinement for Joint Depth and Surface Normal Estimation
Xiaojuan Qi, Zhengzhe Liu, Renjie Liao, Philip H. S. Torr, Raquel Urtasun, Jiaya Jia
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 GeoNet depth-to-normal and normal-to-depth modules (ablation) | VGG-16 baseline · NYUD-V2 normal | mean angle error (deg): 19.4 -> 19.0 | Improved |
| add ensemble module (ablation) | GeoNet · NYUD-V2 normal | mean angle error (deg): 19.0 -> 18.9 | Improved |
| add edge-aware refinement (ablation) | GeoNet · NYUD-V2 normal | mean angle error (deg): 19.0 -> 18.6 | Improved |
| add iterative inference (full model) (ablation) | GeoNet++ · NYUD-V2 normal | mean angle error (deg): 18.6 -> 18.5 | Improved |
| add GeoNet modules (ablation) | VGG-16 baseline · NYUD-V2 depth | rmse: 0.626 -> 0.608 | Improved |
| add full GeoNet++ modules (ablation) | VGG-16 baseline · NYUD-V2 depth | rmse: 0.626 -> 0.600 | Improved |
| add GeoNet++ modules (ablation) | VGG-16 baseline · NYUD-V2 3DGM | mean angle error (deg): 42.39 -> 33.24 | Improved |
| use explicit geometric modules instead of geometric loss | VGG-16 baseline · NYUD-V2 depth | rmse: 0.615 -> 0.600 | Improved |
| change Canny thresholds to extremely low values (ablation) | GeoNet++ · NYUD-V2 normal | mean angle error (deg): 18.6 -> 18.8 | No clear effect |
| change Canny thresholds to extremely high values (ablation) | GeoNet++ · NYUD-V2 depth | rmse: 0.605 -> 0.605 | No clear effect |
| add GeoNet++ refinement on top of DORN | DORN · KITTI depth | rmse: 4.06 -> 4.10 | Hurt |
| add GeoNet++ refinement on top of DORN | DORN · NYUD-V2 depth | rmse: 0.552 -> 0.527 | Improved |
Closest papers by meaning
- GeoNet: Geometric Neural Network for Joint Depth and Surface Normal Estimation (CVPR 2018 · similarity 0.96)
- Unsupervised Learning of Geometry From Videos With Edge-Aware Depth-Normal Consistency (AAAI 2018 · similarity 0.91)
- CORE: Co-planarity Regularized Monocular Geometry Estimation with Weak Supervision (ICCV 2023 · similarity 0.89)
- Adaptive Surface Normal Constraint for Geometric Estimation From Monocular Images (TPAMI 2024 · similarity 0.88)
- Refine-Net: Normal Refinement Neural Network for Noisy Point Clouds (TPAMI 2022 · similarity 0.88)
- Virtual Normal: Enforcing Geometric Constraints for Accurate and Robust Depth Prediction (TPAMI 2021 · 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.
