Python script for deep learning-based extraction of road networks

This dataset includes two Python scripts for extracting road networks from topographic maps or other digital imagery. The Script is designed to automatically classify roads on multiple image files. The first script, “DeepLearning_Training.py”, creates a trained model for deep learning road extractions using ArcGIS Pro’s Multi-Task Road Extractor. The trained model can then be used in the second script, “DeepLearning_Classification.py”, to extract classified rasters using the ArcGIS Pro Classify Pixels Using Deep Learning tool. After classification, the script converts all classified tiles into polygon shapefiles and merges them into a single shapefile. The script has been used to create the historical road network of Kenya, accessible at DOI: 10.5880/TRR228DB.37. Please note: A commercial ArcGIS Pro license is required to execute these scripts.