LAYS: Cross-View Yaw Estimation with Line-Aligning Yaw Scoring

Abstract

Accurate yaw estimation is a bottleneck in cross-view localization between ground view and Bird's Eye View (BEV). Existing methods couple yaw with translation and rely on height or projection assumptions that degrade under large yaw ambiguity. We disentangle yaw from location accuracy and introduce LAYS, a radially invariant line-consensus voting method. By exploiting the radial invariance of our formulation, we achieve sub-degree yaw precision via 3D voting over all candidate poses, while eliminating the need for accurate location.

Our key observation is that a ground-image column matched to BEV pixels induces the same yaw across all camera positions along the radial direction of the pixels. LAYS matches BEV pixels to ground columns using feature similarity and accumulates the induced yaw votes into discrete 3D bins, where correct correspondences along the radial line concentrate into a sharp peak for the correct yaw. Experiments on Mapillary, Ford, KITTI, and VIGOR show significant gains under unknown yaw (+28~45%p for normal FoV), and using LAYS as a yaw prior further improves downstream 3-DoF localization.

Method

Ground–BEV yaw is hard when location is uncertain: pixel-to-pixel matchers entangle orientation with translation and often depend on ground height. LAYS instead treats yaw as a line alignment problem.

Ground pixels are aggregated into column features (height-free). Matching a column to BEV pixels defines a radial line whose induced absolute yaw is the same for every camera position along that ray. Match scores vote into a 3D pose tensor (yaw × 2D location); correct correspondences concentrate at the true yaw while incorrect matches disperse.

Results

Under fully unknown yaw (±180°) and ±20 m location noise, LAYS improves angular accuracy by large margins on Mapillary, Ford, KITTI, and VIGOR (about +28–45%p within normal FoV settings in the paper). Using LAYS as a yaw prior also improves downstream 3-DoF localization.

Quantitative Results

Angular accuracy (% of estimates within θ° of ground truth) under ±20m location noise on four benchmarks. LAYS achieves state-of-the-art results across all settings, with particularly large gains under fully unknown yaw (±180°).

Mapillary Geo-Localization (MGL)

Method	±45° yaw noise			±180° yaw noise
Method	<1°	<2°	<4°	<1°	<2°	<4°
CCVPE	21.36	40.78	68.15	6.55	12.55	23.79
BoostAcc	7.40	14.19	27.54	0.58	1.09	2.13
G2S	32.67	57.49	82.50	0.59	1.27	1.86
FG2	18.59	34.97	59.20	3.13	6.28	12.44
LAYS (Ours)	72.10	90.16	95.63	34.81	52.42	61.74

KITTI (±180° unknown yaw)

Method	Same Area			Cross Area
Method	<1°	<2°	<4°	<1°	<2°	<4°
CCVPE	8.96	≤26.48	≤42.75	3.14	≤9.24	≤14.56
BoostAcc	0.53	0.90	1.96	0.56	1.05	2.09
G2S	0.58	1.17	2.17	0.54	0.82	1.64
FG2	1.62	3.15	6.47	1.62	2.96	6.13
LAYS (Ours)	52.98	65.07	67.72	49.59	58.92	60.25

Ford Multi-AV Highway (±45° yaw noise)

Method	<1°	<2°	<4°
CCVPE	41.36	62.74	75.11
BoostAcc	24.38	43.24	67.86
G2S	16.81	47.71	68.43
FG2	16.75	41.84	57.13
LAYS (Ours)	67.05	85.14	91.76

BibTeX

@inproceedings{kang2026lays,
  author    = {Kang, Taeho and Zhang, Nairan and Kim, Yelin and Shi, Yujiao and Lee, Youngki},
  title     = {Cross-View Yaw Estimation in Location Uncertainty with Line-Aligning Yaw Scoring},
  booktitle = {Computer Vision - {ECCV} 2026 - 19th European Conference, Malm{\"{o}}, Sweden, September 8-12, 2026, Proceedings},
  year      = {2026},
  month     = {September},
  publisher = {Springer},
  series    = {Lecture Notes in Computer Science},
}