UrbanTree3D Dataset
Authors/Creators
- 1. Research Unit of Geospatial Technologies for a Smart Decision, Hassan II Institute of Agronomy and Veterinary Medicine. 10101 Rabat, Morocco
- 2. Société Topographie Informatique, 91000 Evry Courcouronnes, France
- 3. School of Geomatics and Surveying Engineering. Hassan II Institute of Agronomy and Veterinary Medicine. 10101 Rabat, Morocco
- 4. Research Unit of Geospatial Technologies for a Smart Decision, Hassan II Institute of Agronomy and Veterinary Medicine.
- 5. Department of Photogrammetry and Cartography, School of Geomatics and Surveying Engineering
- 6. Department of Applied Statistics and Computer Science
Description
The UrbanTree3D dataset consists of 152 individual urban trees, generated through the integration of airborne LiDAR data, field inventory records, and in-situ field verification. This multi-source approach ensures a high level of annotation reliability and spatial accuracy, making the dataset suitable for advanced machine learning and deep learning applications.
Following preprocessing and data cleaning procedures, six distinct urban tree species were retained, providing a representative diversity for classification tasks in complex urban environments.
The UrbanTree3D includes the following species:
- (0) Aesculus hippocastanum : 62 trees
- (1) Prunus serrulata : 12 trees
- (2) Pyrus calleryana ‘Chanticleer’ : 15 trees
- (3) Salix alba : 23 trees
- (4) Tilia platyphyllos : 25 trees
- (5) Tilia × europaea ‘Pallida’ : 15 trees
UrbanTree3D is designed to support a wide range of applications, including urban tree species classification, advanced 3D point cloud analysis, and the benchmarking of deep learning models applied to urban vegetation. It also serves as a valuable resource for research in urban ecology and smart city applications, enabling more accurate and data-driven management of urban green infrastructure.