PRUE: A Practical Recipe for Field Boundary Segmentation at Scale

2026-03-27

AI Summary

This study addresses the limited generalization of existing satellite-based agricultural field boundary segmentation methods under variations in illumination, scale, and geography, which hinders global mapping efforts. For the first time, it systematically evaluates 18 segmentation and geospatial foundation models under a unified experimental protocol and introduces a robust segmentation framework based on a U-Net architecture, a composite loss function, and tailored data augmentation strategies. Evaluated on the Fields of The World (FTW) benchmark, the proposed method achieves 76% IoU and 47% object-F1, representing improvements of 6% and 9% over previous baselines, respectively, and demonstrates substantially enhanced cross-regional generalization. The project also releases the trained models and a high-quality field boundary dataset covering five countries.

Abstract

Large-scale maps of field boundaries are essential for agricultural monitoring tasks. Existing deep learning approaches for satellite-based field mapping are sensitive to illumination, spatial scale, and changes in geographic location. We conduct the first systematic evaluation of segmentation and geospatial foundation models (GFMs) for global field boundary delineation using the Fields of The World (FTW) benchmark. We evaluate 18 models under unified experimental settings, showing that a U-Net semantic segmentation model outperforms instance-based and GFM alternatives on a suite of performance and deployment metrics. We propose a new segmentation approach that combines a U-Net backbone, composite loss functions, and targeted data augmentations to enhance performance and robustness under real-world conditions. Our model achieves a 76\% IoU and 47\% object-F1 on FTW, an increase of 6\% and 9\% over the previous baseline. Our approach provides a practical framework for reliable, scalable, and reproducible field boundary delineation across model design, training, and inference. We release all models and model-derived field boundary datasets for five countries.

Problem

• field boundary segmentation
• satellite imagery
• geospatial foundation models
• agricultural monitoring
• large-scale mapping

Innovation

• field boundary segmentation
• U-Net
• geospatial foundation models
• composite loss
• data augmentation