MapGCLR: Geospatial Contrastive Learning of Representations for Online Vectorized HD Map Construction

Jonas Merkert; Alexander Blumberg; Jan-Hendrik Pauls; Christoph Stiller

cs.ROMar 11, 2026

MapGCLR: Geospatial Contrastive Learning of Representations for Online Vectorized HD Map Construction

Authors: Jonas Merkert, Alexander Blumberg, Jan-Hendrik Pauls, Christoph Stiller

Organizations: Institute of Measurement and Control Systems, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany

Abstract

Autonomous vehicles rely on map information to understand the world around them. However, the creation and maintenance of offline high-definition (HD) maps remains costly. A more scalable alternative lies in online HD map construction, which only requires map annotations at training time. To further reduce the need for annotating vast training labels, self-supervised training provides an alternative. This work focuses on improving the latent birds-eye-view (BEV) feature grid representation within a vectorized online HD map construction model by enforcing geospatial consistency between overlapping BEV feature grids as part of a contrastive loss function. To ensure geospatial overlap for contrastive pairs, we introduce an approach to analyze the overlap between traversals within a given dataset and generate subsidiary dataset splits following adjustable multi-traversal requirements. We train the same model supervised using a reduced set of single-traversal labeled data and self-supervised on a broader unlabeled set of data following our multi-traversal requirements, effectively implementing a semi-supervised approach. Our approach outperforms the supervised baseline across the board, both quantitatively in terms of the downstream tasks vectorized map perception performance and qualitatively in terms of segmentation in the principal component analysis (PCA) visualization of the BEV feature space.

Figures & tables

Fig. 1: Overlapping BEV feature grids visualized using PCA. In the background, we depict ground truth map labels with road boundaries, lane dividers, centerlines and pedestrian crossing. Both the geospatial consistency between overlapping BEV feature spaces as well as their alignment with the target road layout can be recognized.

Fig. 2: Schematic overview of the semi-supervised learning pipeline. Data flows are shown for supervised (pink) and self-supervised (blue, orange) samples. $\mathcal{B}$ denotes birds-eye-view grids with cell features $\mathbf{f}$ and embeddings $\mathbf{z}$ produced by projection head $h$ . Losses $\mathcal{L}_{\text{sup}}$ and $\mathcal{L}_{\text{GCLR}}$ correspond to supervised and self-supervised branches, respectively.

Fig. 3: Single- and multi-traversals within Argoverse 2: The histogram on the left shows the general distribution of intersecting drive logs. On the right, we visualize the geospatial distribution of single- (red) and multi-traversals (blue) within Miami.

Sup. Data	SSL	AP ${}_{\text{dsh}}$	AP ${}_{\text{sol}}$	AP ${}_{\text{bou}}$	AP ${}_{\text{cen}}$	AP ${}_{\text{ped}}$	mAP	$\Delta_{\mathrm{abs}}$	$\Delta_{\mathrm{rel}}$
2.5%	✗	4.3	5.0	9.6	11.9	1.5	6.5	+2.0	+31%
2.5%	✓	5.2	6.7	12.2	17.0	1.6	8.5	+2.0	+31%
5%	✗	10.3	9.5	20.5	19.1	7.3	13.3	+5.6	+42%
5%	✓	15.4	18.7	24.8	25.4	9.9	18.9	+5.6	+42%
10%	✗	17.6	20.9	31.9	27.1	12.4	22.0	+5.3	+24%
10%	✓	20.8	30.5	34.5	32.4	18.2	27.3	+5.3	+24%

TABLE I: Comparison of the supervised and semi-supervised training strategies for the categories divider dashed (dsh), solid (solid), boundary (bou), centerline (cen) and pedestrian crossing (ped). The first two columns denote the share of supervised training data as well as the use of self-supervised learning (SSL).

Panel of Figure 5 — Fig. 4: Performance scaling and relative gains across increasing share of supervised data.

Fig. 5: Qualitative visualization of PCA of the supervised baseline (middle) and our semi-supervised approach (right) with ground-truth labels in the background for reference. Input surround view on the left, 20 % supervised split.

HD Map Construction

High-definition map construction turns sensor and geographic data into precise, structured road maps for autonomous driving and simulation.

See all 39 papers

GaussianMap: Learning Gaussian Representation for Multi-Sensor Online HD Map Construction
Jun 30, 2026Hongyu Lyu, Julie Stephany Berrio Perez, Mao Shan +1Representation Learning HD Map Construction
The University of Sydney, Australian Centre for Robotics, Sydney, Australia. · The University of Queensland, Australia.
MapTCL: Temporal Consistency Learning via Bidirectional Alignment for Vectorized HD Map Construction
Aug 5, 2026Hyeonseo Kim, Juyeb Shin, Hyeonjun Jeong +2HD Map Construction Semantic Mapping
Robotics Program, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, South Korea · Cho Chun Shik Graduate School of Mobility, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, South Korea
MapLightning: Online Vectorized HD Map Construction with 1D Map Tokens
Oct 1, 2026Shen Zheng, Anurag Ghosh, Mani Ramanagopal +1HD Map Construction
Carnegie Mellon University

Jun 30, 2026J/K move · Enter open · S save

GaussianMap: Learning Gaussian Representation for Multi-Sensor Online HD Map Construction

Hongyu Lyu, Julie Stephany Berrio Perez, Mao Shan, Stewart Worrall

The University of Sydney, Australian Centre for Robotics, Sydney, Australia. · The University of Queensland, Australia.

Representation Learning HD Map Construction

Open paper AskPDF arXiv

Autonomous driving systems benefit from high-definition (HD) maps that provide critical information about road infrastructure. The online construction of HD maps offers a scalable approach to generate local vectorized maps from onboard sensor observations. Existing methods commonly adopt bird's-eye-view (BEV) features as the intermediate scene representation, encoding the surrounding space with fixed-resolution dense grids. However, map elements are spatially sparse yet require fine-grained geometric localization, making uniformly allocated BEV representations redundant and less effective for vectorized map prediction. In this work, we propose GaussianMap, an online HD map construction framework that learns an adaptive Gaussian representation of the surrounding scene. This representation consists of a set of Gaussian primitives on the BEV plane, each encoding a flexible local region with geometric properties and a feature vector, allowing the model to allocate representational capacity to map-relevant regions. To generate such a representation from sensor observations, we introduce a feed-forward Gaussian encoder that progressively refines these primitives through Gaussian interaction modeling and multi-sensor feature aggregation. The refined Gaussian representation is then splatted into a BEV feature map and decoded into vectorized map predictions. Extensive experiments on nuScenes and Argoverse 2 datasets demonstrate that GaussianMap achieves state-of-the-art performance in both camera-only and camera-LiDAR fusion settings. Our code will be made publicly available.

Abstract

HD Map Construction

Explore similar work

GaussianMap: Learning Gaussian Representation for Multi-Sensor Online HD Map Construction

MapTCL: Temporal Consistency Learning via Bidirectional Alignment for Vectorized HD Map Construction

MapLightning: Online Vectorized HD Map Construction with 1D Map Tokens