mlsanityData to 7 Oct 2026

A Data-Centric Approach to 3D Semantic Segmentation of Railway Scenes

arXiv 2504.18213 ↗First posted 25 Apr 2025cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

LiDAR-based semantic segmentation is critical for autonomous trains, requiring accurate predictions across varying distances. This paper introduces two targeted data augmentation methods designed to improve segmentation performance on the railway-specific OSDaR23 dataset. The person instance pasting method enhances segmentation of pedestrians at distant ranges by injecting realistic variations into the dataset. The track sparsification method redistributes point density in LiDAR scans, improving track segmentation at far distances with minimal impact on close-range accuracy. Both methods are evaluated using a state-of-the-art 3D semantic segmentation network, demonstrating significant improvements in distant-range performance while maintaining robustness in close-range predictions. We establish the first 3D semantic segmentation benchmark for OSDaR23, demonstrating the potential of data-centric approaches to address railway-specific challenges in autonomous train perception.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 24 of its 40 references that are in the map. First posted on arXiv 25 Apr 2025.

Builds on

24 of 40 references.

References · 24

  1. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,223
  2. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  3. 2019Scalability in Perception for Autonomous Driving: Waymo Open Datasetcited by 2,436
  4. 2015SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentationcited by 1,483
  5. 2020Point Transformercited by 1,462
  6. 2019KPConv: Flexible and Deformable Convolution for Point Cloudscited by 1,202
  7. 2018PointRCNN: 3D Object Proposal Generation and Detection from Point Cloudcited by 939
  8. 2022BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representationcited by 844
  9. 2017Frustum PointNets for 3D Object Detection from RGB-D Datacited by 640
  10. 2020Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentationcited by 388
  11. 2022Point Transformer V2: Grouped Vector Attention and Partition-based Poolingcited by 384
  12. 2021RPVNet: A Deep and Efficient Range-Point-Voxel Fusion Network for LiDAR Point Cloud Segmentationcited by 173
12 more
  1. 2022Point-to-Voxel Knowledge Distillation for LiDAR Semantic Segmentationcited by 118
  2. 2020EfficientPS: Efficient Panoptic Segmentationcited by 111
  3. 2021EfficientLPS: Efficient LiDAR Panoptic Segmentationcited by 76
  4. 2022PolarMix: A General Data Augmentation Technique for LiDAR Point Cloudscited by 74
  5. 2018Attention-guided Unified Network for Panoptic Segmentationcited by 71
  6. 2022CENet: Toward Concise and Efficient LiDAR Semantic Segmentation for Autonomous Drivingcited by 56
  7. 2020Improving 3D Object Detection through Progressive Population Based Augmentationcited by 41
  8. 2020Part-Aware Data Augmentation for 3D Object Detection in Point Cloudcited by 34
  9. 2020Are we Missing Confidence in Pseudo-LiDAR Methods for Monocular 3D Object Detection?cited by 27
  10. 2023Advancements in Point Cloud Data Augmentation for Deep Learning: A Surveycited by 24
  11. 2023OSDaR23: Open Sensor Data for Rail 2023cited by 17
  12. 2023MemorySeg: Online LiDAR Semantic Segmentation with a Latent Memorycited by 13

Built on by

Not cited yet.