Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Turin3D: Evaluating Adaptation Strategies under Label Scarcity in Urban LiDAR Segmentation with Semi-Supervised Techniques

Luca Barco, Giacomo Blanco, Gaetano Chiriaco, Alessia Intini, Luigi La Riccia, Vittorio Scolamiero, Piero Boccardo, Paolo Garza, Fabrizio Dominici

TL;DR

We address label scarcity in urban 3D semantic segmentation by introducing Turin3D, a high-density aerial LiDAR dataset over Turin with unlabeled training data and manually annotated validation/test sets. The study benchmarks multiple architectures under transfer learning and proposes an iterative semi-supervised pseudo-labeling strategy, demonstrating that adaptive confidence thresholds yield substantial gains over a transfer-learning baseline. RandLA-Net emerges as the most effective backbone for cross-city adaptation, and the adaptive semi-supervised approach achieves an absolute improvement of $9.76$ mIoU, highlighting the value of unlabeled data when ground truth is scarce. The dataset and findings aim to promote data-efficient learning and domain adaptation for large-scale urban LiDAR segmentation, with potential extensions in annotation, taxonomy, and modality fusion.

Abstract

3D semantic segmentation plays a critical role in urban modelling, enabling detailed understanding and mapping of city environments. In this paper, we introduce Turin3D: a new aerial LiDAR dataset for point cloud semantic segmentation covering an area of around 1.43 km2 in the city centre of Turin with almost 70M points. We describe the data collection process and compare Turin3D with others previously proposed in the literature. We did not fully annotate the dataset due to the complexity and time-consuming nature of the process; however, a manual annotation process was performed on the validation and test sets, to enable a reliable evaluation of the proposed techniques. We first benchmark the performances of several point cloud semantic segmentation models, trained on the existing datasets, when tested on Turin3D, and then improve their performances by applying a semi-supervised learning technique leveraging the unlabelled training set. The dataset will be publicly available to support research in outdoor point cloud segmentation, with particular relevance for self-supervised and semi-supervised learning approaches given the absence of ground truth annotations for the training set.

Turin3D: Evaluating Adaptation Strategies under Label Scarcity in Urban LiDAR Segmentation with Semi-Supervised Techniques

TL;DR

We address label scarcity in urban 3D semantic segmentation by introducing Turin3D, a high-density aerial LiDAR dataset over Turin with unlabeled training data and manually annotated validation/test sets. The study benchmarks multiple architectures under transfer learning and proposes an iterative semi-supervised pseudo-labeling strategy, demonstrating that adaptive confidence thresholds yield substantial gains over a transfer-learning baseline. RandLA-Net emerges as the most effective backbone for cross-city adaptation, and the adaptive semi-supervised approach achieves an absolute improvement of mIoU, highlighting the value of unlabeled data when ground truth is scarce. The dataset and findings aim to promote data-efficient learning and domain adaptation for large-scale urban LiDAR segmentation, with potential extensions in annotation, taxonomy, and modality fusion.

Abstract

3D semantic segmentation plays a critical role in urban modelling, enabling detailed understanding and mapping of city environments. In this paper, we introduce Turin3D: a new aerial LiDAR dataset for point cloud semantic segmentation covering an area of around 1.43 km2 in the city centre of Turin with almost 70M points. We describe the data collection process and compare Turin3D with others previously proposed in the literature. We did not fully annotate the dataset due to the complexity and time-consuming nature of the process; however, a manual annotation process was performed on the validation and test sets, to enable a reliable evaluation of the proposed techniques. We first benchmark the performances of several point cloud semantic segmentation models, trained on the existing datasets, when tested on Turin3D, and then improve their performances by applying a semi-supervised learning technique leveraging the unlabelled training set. The dataset will be publicly available to support research in outdoor point cloud segmentation, with particular relevance for self-supervised and semi-supervised learning approaches given the absence of ground truth annotations for the training set.

Paper Structure

This paper contains 23 sections, 1 equation, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Datasets for Urban Mapping
  4. UAV-Based Datasets
  5. Aerial Datasets
  6. Mobile Datasets
  7. 3D Semantic Segmentation
  8. Turin3D Dataset
  9. LiDAR Acquisition
  10. 3D Point Cloud Processing
  11. Dataset Description
  12. Semantic Labels Taxonomy
  13. Annotation Process
  14. Methodology
  15. Problem Formulation
  16. ...and 8 more sections

Figures (3)

  • Figure 1: Turin3D point cloud whole extent and subdivision in blocks
  • Figure 2: Close-in views of Turin3D. Top row displays the scenes in RGB coloring, bottom row shows the same areas with points colored according to their assigned class labels.
  • Figure 3: Distribution of classes across test and validation sets. The percentage indicates the proportion of each of the six classes within their respective sets, not relative to the entire dataset.