Tools
Tools
Finlay, Mark (2025) TrailSurfNET: Trail Surface Classification Using Convolutional Neural Networks and OpenStreetMap Annotations. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (2MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (3MB) | Preview |
Abstract
Accurate classification of hiking trail surfaces is essential for improving outdoor navigation, enhancing safety, and supporting effective land management practices. This research addresses significant data gaps in OpenStreetMap (OSM), where surface-type labels are often incomplete or inconsistent. It introduces and evaluates TrailSurfNET, a novel framework that integrates multi-band Sentinel-2 satellite imagery with OSM annotations to automatically classify trail surfaces.
The study develops a scalable data pipeline that harmonizes diverse OSM tags into five distinct classes (asphalt, paved, gravel, mud/dirt, grass) and generates a balanced dataset through targeted down-sampling. It then conducts a comparative analysis of multiple Convolutional Neural Network (CNN) architectures, including VGG, ResNet-18, ResNet34, and ResNet-50. The ResNet were evaluated under two conditions: trained from scratch, and fine-tuned with weights pre-trained on the BigEarthNet satellite imagery dataset.
Two key findings emerged: (1) training from scratch matched or exceeded domain-specific transfer learning, and (2) deeper ResNet variants offered a small accuracy gain; the best model (ResNet-50, from scratch) reached 46.73% OA. The findings confirm the viability of using deep learning to augment OSM, providing a scalable solution to enrich critical trail metadata and enhance route planning applications for sustainable trail management.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Onwuegbuche, Faithful Chiagoziem UNSPECIFIED |
| Subjects: | Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning |
| Divisions: | School of Computing > Master of Science in Artificial Intelligence |
| Depositing User: | Ciara O'Brien |
| Date Deposited: | 28 May 2026 13:49 |
| Last Modified: | 28 May 2026 13:49 |
| URI: | https://norma.ncirl.ie/id/eprint/9319 |
Actions (login required)
 |
View Item |