FIP 1.0 soybean data: Insights on soybean growth from eight years of high-throughput image field phenotyping

Why watching crops from above matters

As climate extremes become more common, farmers and breeders need crops that keep growing well in good years and bad. This study introduces an unusually rich, eight-year record of how soybean plants grew in a real farm field, captured by a camera system hanging over the crop. By linking millions of visible leaves to detailed weather records, the dataset helps scientists understand which soybean types cope best with heat, drought, or cloudy spells, and gives modelers the raw material to predict future harvests more reliably.

A field turned into a living laboratory

From 2015 to 2022, researchers in Switzerland grew 72 soybean varieties and breeding lines in carefully laid-out plots. Above them, a rope-suspended platform carried a high-resolution camera about three meters over the plants. Throughout each growing season, this system took thousands of top-down color photos that clearly captured the crop canopy – the green layer formed by leaves covering the soil. At the same site, a nearby weather station recorded temperature, light, rainfall, humidity, and wind every hour, creating a synchronized picture of plants and their environment.

Turning images into growth curves

To turn raw photos into useful measurements, the team built an image-processing pipeline. First, each picture was filtered so that green pixels belonging to plant leaves were separated from the brown soil background. Next, the algorithm detected the crop rows, corrected for camera tilt, and outlined the exact area belonging to each plot. From there, the researchers calculated “canopy cover” – the fraction of each plot’s area that was green at a given time. Because images were captured regularly through the season, these values trace smooth curves that show when plants emerged, how fast they filled the ground, when they reached peak cover, and when they began to yellow and thin out.

Connecting growth, weather, and yield

The image-based growth records were paired with traditional harvest data such as grain yield, protein content, seed weight, and leaf color readings. The scientists used statistical tools to check how consistent the image traits were across repeated plots and over time. They found that canopy cover and plant height, measured from laser scans in two years, were highly “heritable” – differences between soybean types were strong and repeatable. By fitting smooth curves to canopy cover, they could pinpoint when each variety lost half of its green cover, a marker of leaf aging. Across years, this timing matched known maturity groups and in many cases was clearly linked to final yield: varieties that kept their canopy green for longer often produced more grain, unless weather conditions cut their season short.

A reusable toolbox for crop science

The final dataset gathers more than 17,000 images, their segmented masks, plot layouts, daily canopy cover values, genotype summaries, and matched weather files in simple, widely used formats. It is openly available through institutional and machine-learning repositories, and follows community standards for naming and metadata so that it can be combined with other field imaging efforts, including drone flights over crops. The authors also share the code used to extract canopy cover, inviting others to apply the same workflow to their own experiments or to train new artificial-intelligence models on plant images.

What this means for future harvests

For non-specialists, the key message is that crop breeding is moving from occasional field notes to continuous “movies” of plant life. By watching soybean plots grow leaf by leaf, under real weather and farm conditions, researchers can spot which lines stay vigorous under stress and feed that knowledge into better crop models. The FIP 1.0 soybean dataset does not by itself guarantee higher yields, but it supplies the detailed evidence needed to choose stronger varieties and to design farming practices that keep soybean production stable in an uncertain climate.

Citation: Keller, B., Kirchgessner, N., Oppliger, C. et al. FIP 1.0 soybean data: Insights on soybean growth from eight years of high-throughput image field phenotyping. Sci Data 13, 476 (2026). https://doi.org/10.1038/s41597-026-06663-z

Keywords: soybean phenotyping, crop imaging, canopy cover, climate-resilient crops, field datasets