- Dataset Overview
- Study Area
- Acquisition Platform
- Data Modalities and Features
- Semantic Classes
- Dataset Access and Preview
- Dataset Directory Structure
- License
- Acknowledgments
| Property | Value |
|---|---|
| Dataset Name | ITU-Campus3D |
| Acquisition Method | Mobile Laser Scanning (MLS) |
| Environment Type | Campus Environment |
| Number of Classes | 10 |
| Primary Task | 3D Semantic Segmentation |
| Data Modalities | XYZ, RGB, Intensity, Semantic Labels |
The study area contains a diverse set of urban structures and landscape elements, providing substantial structural variability for semantic scene understanding tasks.
| Property | Details |
|---|---|
| Location | Ayazağa Campus, Istanbul Technical University (ITU) |
| Coordinate System | ETRS89 / UTM Zone 35N (EPSG:3047) |
| Key Features | Road surfaces, sidewalks, vegetation, buildings, traffic infrastructure, and vehicles. |
The mobile mapping platform integrates multiple sensors to ensure accurate georeferencing, trajectory estimation, and high-fidelity colorization of 3D point clouds.
| Sensor Type | Model / Details | Purpose |
|---|---|---|
| LiDAR | Velodyne HDL-32E | 3D point cloud acquisition |
| GNSS | Dual GNSS receivers | Global positioning |
| IMU | Inertial Measurement Unit | Trajectory estimation |
| Odometry | Wheel encoder | Motion estimation |
| Camera | Ladybug 5+ panoramic camera | 3D point cloud colorization |
These features enable both traditional geometric analysis and deep learning-based semantic segmentation workflows.
| Attribute | Description |
|---|---|
| XYZ | Geographical coordinates |
| RGB | RGB color information from panoramic imagery |
| Intensity | LiDAR intensity |
| Semantic Label | Ground-truth semantic classes |
The dataset is categorized into 10 semantic classes, addressing a wide spectrum of urban objects:
| ID | Class | Description |
|---|---|---|
0 |
Ground | Soil, natural earth, and non-road natural surfaces. |
1 |
Road | Vehicle lanes, cycling lanes, and asphalt road surfaces. |
2 |
Road markings | Lane lines, arrows, crosswalks, stop lines, and surface markings. |
3 |
Sidewalk | Pedestrian walking areas and paved sidewalks. |
4 |
Vegetation | Trees, bushes, grass, and general plant life. |
5 |
Building | Building facades, walls, and permanent structural elements. |
6 |
Pole | Light poles, traffic signal poles, and similar vertical supports. |
7 |
Sign | Informational panels and billboards. |
8 |
Vehicle | Cars, vans, and similar motorized vehicles. |
9 |
Other | Objects not belonging to any predefined semantic class, such as vase, bin, and bench. |
To request access, please fill out the form below:
Orthogonal visualization of a representative labeled scene from the ITU-Campus3D dataset.
The dataset will be provided in one of the directory structures below.
ITU-Campus3D/
├── train/
│ └── 0_13.las # Training tiles (*.las)
├── val/
│ └── 57_20.las # Validation tiles (*.las)
├── test/
│ └── 50_12.las # Testing tiles (*.las)
└── buffer/
└── 48_12.las # Buffer tiles (*.las)
If you use ITU-Campus3D in your research, please cite our paper:
@ARTICLE{11477876,
author={Aksu, Koray and Demirel, Hande and Alkan, Reha Metin and Yanalak, Mustafa},
journal={IEEE Access},
title={ITU-Campus3D: A Campus-Scale Urban Mobile Mapping Benchmark for Semantic Segmentation},
year={2026},
volume={14},
number={},
pages={55621-55640},
keywords={3D point cloud;campus environment;class imbalance;digital twin;mobile laser scanning;semantic segmentation;urban scene understanding},
doi={10.1109/ACCESS.2026.3682205}}
The ITU-Campus3D dataset is released under the Creative Commons Attribution 4.0 (CC BY 4.0).
This work was supported by Istanbul Technical University Scientific Research Projects Coordination Unit (Production of High-Definition Maps (HD Maps) for Istanbul Technical University Ayazaga Campus Digital Road Infrastructure under Project MGA-2022-43340).