High-Resolution Road Surface Cartographic Product for 15 Countries in West Africa (WARE-Net)

This repository provides the first comprehensive, high-resolution road surface cartographic product covering 286 cities across 15 West African countries. Generated using the state-of-the-art WARE-Net deep learning model, this product explicitly addresses the severe "digital divide" in infrastructural mapping in the Global South. The source code of proposed WARE-Net is available at github.com.

If If you use WARE-Net or the WARE dataset in your research, please cite:

@article{Lei2026WARE,
title = {WARE-Net: Visual Foundation Model-Based Urban Road Extraction in West Africa from High-Resolution Remote Sensing Imagery},
author = {Lei, Qi and Guo, Jianhua and Sun, Xin and Zhang, Rui and Liang, Dong and Sun, Zhongchang and Guo, Huadong},
journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
year = {2026},
note = {Under review}
}