Fields of The World: A Field Guide for Extracting Agricultural Field Boundaries

2026-02-08

AI Summary

This study addresses the critical scarcity of high-precision, large-scale agricultural field boundary data, which severely limits applications such as crop monitoring and yield estimation. The authors present the first global benchmark dataset of field boundaries, encompassing 1.6 million field polygons across 24 countries. They integrate MOSAIKS random convolutional features, deep learning-based segmentation models, and a cloud-optimized geospatial processing pipeline, and provide a command-line inference tool to enable field-level field delineation and crop classification. Under label-limited conditions, the approach achieves macro F1 scores of 0.65–0.75 for crop classification. Precomputed results covering 4.76 million square kilometers across five countries are released, with median field sizes ranging from 0.06 to 0.28 hectares, marking the first field-level mapping that jointly incorporates forest loss attribution and crop type.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Humans and AI: Crowd Sourcing and Human Computation

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Field boundary maps are a building block for agricultural data products and support crop monitoring, yield estimation, and disease estimation. This tutorial presents the Fields of The World (FTW) ecosystem: a benchmark of 1.6M field polygons across 24 countries, pre-trained segmentation models, and command-line inference tools. We provide two notebooks that cover (1) local-scale field boundary extraction with crop classification and forest loss attribution, and (2) country-scale inference using cloud-optimized data. We use MOSAIKS random convolutional features and FTW derived field boundaries to map crop type at the field level and report macro F1 scores of 0.65--0.75 for crop type classification with limited labels. Finally, we show how to explore pre-computed predictions over five countries (4.76M km\textsuperscript{2}), with median predicted field areas from 0.06 ha (Rwanda) to 0.28 ha (Switzerland).

Problem

• field boundary extraction
• agricultural mapping
• crop monitoring
• benchmark dataset
• field-scale analysis

Innovation

• field boundary extraction
• pre-trained segmentation models
• MOSAIKS
• crop type classification
• cloud-optimized inference