WHU-Road3D: A dataset and benchmark towards fine-grained road surface mapping
Organizations: Sanya Science and Education Innovation Park, Wuhan University of Technology, Sanya 572000, China · School of Artificial Intelligence, Wuhan University of Technology, Wuhan 430070, China · State Key Laboratory of Information Engineering in Surveying, Mapping, and Remote Sensing (LIESMARS), Wuhan University, Wuhan 430079, China · School of Mechanical Engineering, Chongqing Technology and Business University, Chongqing, China ·
Abstract
Accurately extracting and mapping road surface elements is crucial for applications such as high-definition maps, road asset management, and intelligent transportation systems. This study first provides a comprehensive review of the benchmark datasets for extracting and mapping road elements, encompassing overhead-to-perspective data, large-to-local scales, and coarse-to-fine details. Subsequently, we thoroughly review the relevant approaches for extracting road surface elements. However, we argue that existing datasets still require enhancements in semantic richness, data modality diversity, and scenario complexity. To bridge these gaps, this paper presents WHU-Road3D, a large-scale multimodal dataset for fine-grained road surface mapping to address these shortcomings. WHU-Road3D provides approximately 100 km of diverse Chinese highways and urban streets. It features a rich combination of MLS point clouds, associated panoramic images, vehicle trajectories, detailed vectorized annotations, and pointwise instance annotations for road surface elements. The dataset carefully annotates 15 distinct categories of road elements, including common markings like road boundaries, zebra crossings, lane dividers, stop lines, as well as 10 types of direction symbols. To our knowledge, WHU-Road3D stands as the first multimodal dataset for mapping such a comprehensive range of road surface elements, designed to facilitate cutting-edge research in instance segmentation, full-element extraction and mapping on road surfaces. Beyond these immediate applications, WHU-Road3D also provides a valuable source for hierarchical road topology reasoning, map updating, and street scene generation. The dataset, along with the code for our benchmark methods, will be made publicly available at https://github.com/WHU-USI3DV/WHU-Road3D.git.
Figures & tables
| LoD | Dataset and Reference | Scene | View and sensor | Coverage | Concerned Category |
| LoD0 | CasNet [ 5 ] | urban area | satellite imagery | 112 km 2 | road regions |
| RoadTracer [ 3 ] | urban area | satellite imagery | 15 cities | road centerline | |
| SpaceNet [ 16 ] | urban area | satellite imagery | 8676.6 km | road centerline | |
| Grab-posis [ 17 ] | urban area | GPS trajectory | 1.0*10 6 km | road map inference, traffic forecast | |
| Sat2Graph [ 18 ] | urban area | satellite imagery | 20 cities | road centerline | |
| MUNO21 [ 19 ] | urban area | satellite imagery | 21 cities | road centerline update |
| District | Road length (km) | Annotation statistics | Scenario | ||||||
| Road boundary (km) | Lane divider (km) | Crosswalk (count) | Stop line (count) | Direction symbol (count) | |||||
| Dashed | Solid | Outline | Stripe | ||||||
| SH | 35.153 | 51.407 | 36.860 | 30.156 | 163 | 2447 | 168 | 633 | Urban street + expressway |
| WH | 26.180 | 23.300 | 22.797 | 34.490 | 88 | 1195 | 51 | 438 | Urban street |
| CM | 34.185 | 32.782 | 73.779 | 47.178 | 65 | 1350 | 71 | 511 | Urban street |
| Category | District | ||
|---|---|---|---|
| SH | WH | CM | |
| Forward arrow | 238 | 263 | 267 |
| Left arrow | 141 | 48 | 36 |
| Right arrow | 54 | 20 | 18 |
| Forward & left arrow | 37 | 39 | 6 |
| Forward & right arrow | 128 | 62 | 39 |
| Method | Category | Precision | Recall | F1 | |
| MapTR [ 76 ] | Solid lane divider | 0.731 | 0.723 | 0.727 | 1.738 |
| Dashed lane divider | 0.752 | 0.795 | 0.773 | 1.763 | |
| Road boundary | 0.397 | 0.584 | 0.473 | 3.356 | |
| Crosswalk outline | 0.364 | 0.244 | 0.292 | 4.232 | |
| WHU-lane [ 45 ] | Solid lane divider | 0.821 | 0.783 | 0.801 | 1.515 |
| Dashed lane divider | 0.794 | 0.895 | 0.842 | 2.214 |
| Category | TP | FP | FN | GT | Precision | Recall | F1 | |
|---|---|---|---|---|---|---|---|---|
| Forward arrow | 251 | 25 | 12 | 263 | 0.909 | 0.954 | 0.931 | 0.623 |
| Left arrow | 44 | 3 | 4 | 48 | 0.936 | 0.917 | 0.926 | 0.725 |
| Right arrow | 18 | 3 | 2 | 20 | 0.857 | 0.900 | 0.878 | 0.708 |
| Forward & left arrow | 37 | 4 | 2 | 39 | 0.902 | 0.949 | 0.925 | 0.697 |
| Forward & right arrow | 54 | 3 | 8 | 62 | 0.947 | 0.871 | 0.907 | 0.684 |
| Rhombus | 6 | 2 | 0 | 6 | 0.750 | 1.000 | 0.857 | 0.437 |