The scarcity and high cost of labeled high-resolution satellite imagery pose a major challenge to remote sensing applications, especially in low-income regions, where such data are often inaccessible. In this study, we propose a weakly supervised learning approach to estimate parking lot occupancy using low-cost, 3-meter resolution satellite imagery. Using coarse temporal labels based on the cultural assumption that parking lots of major supermarkets and hardware stores in Germany are typically fuller(occupied) on Saturdays and emptier on Sundays - we train a pairwise comparison model that achieves an AUC of 0.92 on large parking lots. To demonstrate real-world applicability, we extended our method to monitor mobility patterns at a major bus terminal in Sudan, successfully capturing a drop in activity during the onset of armed conflict. By reducing reliance on expensive high-resolution imagery, our approach offers a scalable solution for equitable mobility estimation. Moreover, the method can be adapted to assess transit patterns and resource allocation in vulnerable communities, providing a data-driven basis to improve the well-being of those most in need.

PDF