Research map / 3D vision / Depth and stereo
Neural RGB(r)D Sensing: Depth and Uncertainty From a Video Camera
Check 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 |
|---|---|---|---|
| estimate depth probability volumes and integrate them over video frames with adaptive K-Net damping | Neural RGB→D Sensing · 7-Scenes | σ<1.25: 60.05 -> 69.26 | Helped |
| estimate depth probability volumes and integrate them over video frames with adaptive K-Net damping | Neural RGB→D Sensing · KITTI | rmse: 3.1375 -> 2.8294 | Helped |
| estimate depth probability volumes and integrate them over video frames with adaptive K-Net damping | Neural RGB→D Sensing · KITTI | abs. rel: 0.0874 -> 0.0998 | Hurt |
| use the proposed video method rather than DORN in KITTI-to-virtual-KITTI transfer | Neural RGB→D Sensing · KITTI (train) → virtual KITTI (test) | rmse: 9.618 -> 6.452 | Helped |
| use the proposed video method rather than DORN in KITTI-to-virtual-KITTI transfer | Neural RGB→D Sensing · KITTI (train) → virtual KITTI (test) | abs. rel: 0.2256 -> 0.2537 | Hurt |
| use the proposed video method rather than DORN for indoor-to-KITTI transfer | Neural RGB→D Sensing · Indoor (train) → KITTI (test) | σ<1.25: 25.44 -> 72.96 | Helped |
| replace DSO pose initialization on all frames with DSO initialization only in the first time window (ablation) | Neural RGB→D Sensing · 7-Scenes | σ<1.25: 60.63 -> 62.08 | Helped |
| use IMU relative rotations instead of DSO-initialized rotations for all frames (ablation) | Neural RGB→D Sensing · 7-Scenes | σ<1.25: 60.63 -> 69.26 | Helped |
| use ground-truth camera poses instead of IMU relative rotations (ablation) | Neural RGB→D Sensing · 7-Scenes | σ<1.25: 69.26 -> 70.54 | Helped |
| add image-warping difference as an auxiliary training loss (ablation) | Neural RGB→D Sensing · depth prediction | depth quality | No clear effect |
| add a temporal-consistency post-processing network to per-frame estimates (ablation) | DORN, DeMoN, and MonoDepth comparisons · depth estimation | performance | No clear effect |
| apply Bayesian filtering without damping rather than not integrating DPVs over time (ablation) | Neural RGB→D Sensing · video depth estimation | depth noise | Helped |
Closest papers by meaning
- Deep Depth Completion of a Single RGB-D Image (CVPR 2018 · similarity 0.90)
- Single Image Depth Estimation Trained via Depth From Defocus Cues (CVPR 2019 · similarity 0.89)
- Learned Binocular-Encoding Optics for RGBD Imaging Using Joint Stereo and Focus Cues (CVPR 2025 · similarity 0.88)
- Depth Prompting for Sensor-Agnostic Depth Estimation (CVPR 2024 · similarity 0.87)
- RGB-Only Supervised Camera Parameter Optimization in Dynamic Scenes (NeurIPS 2025 · similarity 0.86)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · 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.