Loading...
Unsupervised Urban Tree Biodiversity Mapping from Street Imagery
Abuhani, Diaa Addeen
Abuhani, Diaa Addeen
Date
2025-08
Authors
Advisor
Type
Thesis
Degree
Citations
Altmetric:
Description
A Master of Science thesis in Machine Learning by Diaa Addeen Abuhani entitled, “Unsupervised Urban Tree Biodiversity Mapping from Street Imagery”, submitted in August 2025. Thesis advisor is Dr. Imran Zualkernan and thesis co-advisor is Dr. Martina Mazzarella. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).
Abstract
Urban tree biodiversity plays a critical role in climate resilience, ecological stability, and livability. However, large-scale biodiversity assessments remain limited due to the need for taxonomic labels and expert-led field surveys. In this work, we introduce an unsupervised clustering framework that combines visual embeddings and spatial priors to assess urban tree biodiversity directly from street-level imagery without reliance on labeled data. Our method accurately captures key ecological indicators, particularly Shannon and Simpson entropies, and provides reasonable estimates of species richness across diverse urban contexts. By leveraging the inherent spatial distribution of trees alongside visual features, our approach remains robust to geographic variability and domain shifts. We validate our framework across multiple cities and demonstrate its capacity to recover genus-level biodiversity patterns that align with known ecological distributions. This work provides a scalable pathway for monitoring urban biodiversity and offers a step toward more generalizable, data-efficient ecological assessment tools in support of nature-based urban planning.