This is the official repository of the RailPC dataset. For details, please refer to:
RailPC: A Large-Scale Railway Point Cloud Semantic Segmentation Dataset
Tengping Jiang, Shiwei Li, Qinyu Zhang, Guangshuai Wang, Zequn Zhang, Fankun Zeng, Peng An, Xin Jin, Shan Liu, Yongjun Wang.
Semantic segmentation in the context of 3D point clouds for the railway environment holds significant economic value, but its development is severely hindered by the lack of suitable and specific datasets. Additionally, the models trained on existing urban road point cloud datasets demonstrate poor generalization on railway data due to a large domain gap caused by non-overlapping special/rare categories, e.g., rail track, track bed, etc. To harness the potential of supervised learning methods in the domain of 3D railway semantic segmentation, we introduce RailPC, a new point cloud benchmark. RailPC provides a large-scale dataset with rich annotations for semantic segmentation in the railway environment. Notably, RailPC contains twice the number of annotated points compared to the largest available mobile laser scanning (MLS) point cloud dataset and is the first railway-specific 3D dataset for semantic segmentation. It covers a total of nearly 25 km railway in two different scenes (urban and mountain), with 3 billion points that are finely labeled as 16 most typical classes w.r.t railway, and the data acquisition process is completed in China by MLS systems. Through extensive experimentation, we evaluate the performance of advanced scene understanding methods on the annotated dataset and present a synthetic analysis of semantic segmentation results. Based on our findings, we establish some critical challenges towards railway-scale point cloud semantic segmentation.
The railway point cloud (first part, approximately 7 kilometers) is currently subject to ownership disputes and confidentiality restrictions, and will no longer be publicly accessible. We are in the process of re-labeling the new railway point cloud and aim to complete this labeling by September. Following the completion of the paper's correction and re-labeling process, we will make this dataset publicly available again.
If you would like to apply for this dataset, please complete the information in the following format and email it to [email protected]. We'll get right back to you.
Subject: Request for RailPC dataset
******** Basic information ********
· Name:
· Email Address:
· Affiliation/Unit:
· Contact Phone Number:
******* Research background *******
· Research area/direction:
· How do you plan to use the dataset? (e.g. academic research, business analysis, curriculum projects, etc.):
· Would you be willing to cite or refer to this dataset and its sources in your research results:
· Your suggestions for this dataset(Optional):
The proposed benchmark called the RailPC dataset, it comprises over 2 billion 3D points collected from two different environments (i.e., urban and mountain railway scenes). The urban railway was captured a section of Nanjing, China, covering length of around 10 km. As for data of mountain railway environments, the data were collected in the rural railway yard area of Chengdu. The mountain railway point cloud covers road length of approximately 15 km, which is located in the relative high altitude regions.
The MLS system is equipped with a HiScan-C scanning sensor with 360° field of view, structured light module, a real-time kinematic (RTK) global navigation satellite system (GNSS), an embedded computer and a power supply system. The detection range of the installed laser sensor is 0.5 m – 119 m, ensuring the accuracy of about 2 mm and generating 1,000,000 points per second. The root mean square accuracy is about 0.04 m in a vertical direction, 0.02 m in the horizontal direction, 0.015◦ in the heading direction, and 0.008◦ in pitch and roll directions.
- Rail track
- Track bed (sleeper and ballast)
- Ground(impervious surfaces and rough terrain)
- Roadside
- Catenary (transmission and distribution lines)
- Pull wire
- Guardrail (fences and railway barriers)
- Vegetation (trees and grass)
- Support (bearing cable and positioning tube)
- Pole (power line poles and light poles)
- Transmission tower
- Stair
- Tunnel
- Building (residential, high-rises, and warehouses)
- Concrete
- Other (remaining objects)
- Noisy points (can be deleted)
- 2024/04/23: The dataset is available for download!
- 2024/04/18: Initial release!
- 2024/03/13: The RailPC has been accepted by CAAI!
- Mountain railway scene semantic segmentation dataset！
- Urban railway scene semantic segmentation dataset！
- Railway scene instance segmentation dataset！
- Railway scene panoptic segmentation dataset！
- SPGraph:Large-scale Point Cloud Semantic Segmentation with Superpoint Graphs
- KPConv: Flexible and deformable convolution for point clouds
- RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
- SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds
- Tangent Convolutions for Dense Prediction in 3D
- MS-RRFSegNet: Multi-Scale Regional Relation Feature Segmentation Network for Semantic Segmentation of Urban Scene Point Clouds
- Supervoxel convolution for online 3D semantic segmentation
