Research map / 3D vision / Depth and stereo
Cross-View Yaw Estimation in Location Uncertainty with Line-Aligning Yaw Scoring
Taeho Kang, Nairan Zhang, Yelin Kim, Yujiao Shi, Youngki Lee
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 |
|---|---|---|---|
| use radially invariant line voting instead of point-to-point projection scoring (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 13.19 -> 34.81 | Improved |
| use radially invariant voting instead of polar-transform projective scoring (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 7.59 -> 34.81 | Improved |
| use column-wise feature aggregation instead of keypoint features (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 20.12 -> 34.81 | Improved |
| use weighted-sum column aggregation instead of an MLP (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 31.73 -> 34.81 | Improved |
| use probabilistic column selection in training instead of greedy (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 28.69 -> 34.81 | Improved |
| use hard argmax matching at inference instead of weighted (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 30.77 -> 34.81 | Improved |
| add pose plausibility score S_P (ablation) | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 14.18 -> 34.81 | Improved |
| add pitch and roll noise of +-10 degrees | LAYS · MGL (+-180 deg yaw) | yaw accuracy <1 degree (%): 34.81 -> 33.15 | No clear effect |
| initialize FG2 with LAYS yaw (2-DoF) instead of FG2 two-stage | FG2 · VIGOR cross area | mean localization error (m): 5.95 -> 5.10 | Improved |
| initialize FG2 with LAYS yaw instead of two-stage yaw | FG2 · MGL (+-45 deg) | median localization error (m): 8.95 -> 3.99 | Improved |
| initialize FG2 with LAYS yaw instead of two-stage yaw | FG2 · KITTI (+-180 deg) | median localization error (m): 12.26 -> 1.48 | Improved |
| widen yaw noise from +-45 to +-180 degrees | LAYS · MGL | yaw accuracy <1 degree (%): 72.10 -> 34.81 | Hurt |
Closest papers by meaning
- Seeing Across Skies and Streets: Feedforward 3D Reconstruction from Satellite, Drone, and Ground Images (NeurIPS 2026 · similarity 0.88)
- Sector-Level Cross-View Geo-Localization with Implicit Orientation via Azimuthal Scanning (ECCV 2026 · similarity 0.87)
- BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization (NeurIPS 2025 · similarity 0.87)
- View From Above: Orthogonal-View aware Cross-view Localization (CVPR 2024 · similarity 0.86)
- FG^2: Fine-Grained Cross-View Localization by Fine-Grained Feature Matching (CVPR 2025 · similarity 0.85)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.84)
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.
