Skip to content

Community

Agribound v0.1.3: Query Published FTW Polygons by AOI

Historical post

This post describes agribound 0.1.3 and is kept as it was published, except that Jeremy Rapp's name now links to the ORCID record (the Michigan State directory page it linked to no longer exists). Several statements in it no longer hold for agribound 1.0, and some 0.1.x results were affected by defects fixed in 1.0.0 (see the v1.0.0 release post, the migration guide and the CHANGELOG). The AOI file examples/data/small_aoi.geojson in the examples below is a placeholder for your own AOI; it is not in the repository. For the 1.0 API, see the FTW Polygon Query guide.

Agribound v0.1.3 adds a data-access helper for the published Fields of The World (FTW) global field-boundary polygons. Instead of running FTW inference end-to-end, you can now pull pre-computed FTW predictions for any area of interest with a single function call -- backed by a PyArrow GeoParquet reader against the public Source Cooperative dataset.

This is agribound's second community contribution from Jeremy Rapp at the Department of Earth and Environmental Sciences, Michigan State University -- following his USGS NAIP Plus contribution in v0.1.2.

Introducing Agribound: Unified Field Boundary Delineation from Satellite Imagery

Historical post

This post describes agribound 0.1.x and is kept as it was published, except for these changes made in 1.0.0: a comparison with other tools was removed; the statement that the New Mexico linework "approaches the quality of manual human digitization", for which no evaluation was made, was removed; the gallery sentence now links to the archived 0.1.x gallery and no longer says it covered nine regions, five satellites and all engines (it had ten study areas and no GeoAI example); the manuscript title in the citation was updated (see the note there); and links were updated: the images now point to assets/gallery_0.1x/, the TESSERA post is named and linked through the TESSERA blog index, and Jeremy Rapp's name links to the ORCID record (the Michigan State directory page it linked to no longer exists). Several statements in it no longer hold for agribound 1.0, and some 0.1.x results were affected by defects fixed in 1.0.0 (see the v1.0.0 release post, the migration guide and the CHANGELOG).

We are excited to announce the public release of agribound, a Python package that unifies seven complementary approaches to agricultural field boundary delineation into a single, reproducible pipeline. Whether you are working with 1 m NAIP imagery over U.S. croplands or 10 m Sentinel-2 composites anywhere in the world, agribound lets you go from raw satellite data to clean, vectorized field boundary polygons in a single function call.

Agribound field boundaries over Lea County, NM
Agribound-delineated field boundaries over Lea County, New Mexico (DINOv3 fine-tuned + SAM2 on NAIP). Image credit: Jayden Atkin, DRI.