Pubescence Color Classification in Soybean Breeding Using Aerial Images and the Random Forest Machine Learning Algorithm

Published: 16 September 2025| Version 1 | DOI: 10.17632/f3h24f3m4h.1
Contributors:
,
,

Description

These data comprise a time series of aerial images taken from a drone and field notes of pubescence color from plant rows of the University of Illinois at Urbana-Champaign soybean breeding program. With this information and using the Random Forest algorithm, the fitted models classify gray, light tawny, and tawny pubescence. According to the year of data (2018-2020), the type of images (RGB or multispectral), and whether a single set or a time series of images is considered, each of these ten models can be tested using the code (.R files) and the model weights (.RData files).

Files

Steps to reproduce

With the following directions, you can either reload the model weights (.RData) or rerun the model from the script to confirm reproducibility of 10 models, the results of which will be summarized in a table that we call Table 2 (manuscript under review). Steps to Reproduce the Analysis under the RStudio environment: 1) Download the compressed files (Table 2.zip) • .Rproj — contains the R project that includes the required files to reproduce 10 scenarios: - A single flight date using RGB data (2018), - Four drone flight dates using RGB data (2018), - A single flight date using RGB data (2019), - Four drone flight dates using RGB data (2019), - A single flight date using multispectral data (2019), - Four drone flight dates using multispectral data (2019), - A single flight date using RGB data (2020), - Four drone flight dates using RGB data (2020), - A single flight date using multispectral data (2020), - Four drone flight dates using multispectral data (2020). • .R — each file contains the R code to preprocess data, fit the model, and generate outputs for the above scenarios. • .RData — each file contains the adjusted model object with trained parameters for the above scenarios. 2) Set up the R environment - Install R (≥ 4.4.3 recommended), - Install RStudio (≥ 2024.09.0 recommended). 3) Install the following R packages used in the analysis install.packages(c("caTools", "randomForest", "caret", "dplyr"))

Institutions

  • University of Illinois at Urbana-Champaign
  • Instituto Nacional de Investigacion Agropecuaria

Categories

High Throughput Analysis, Plant Breeding

Licence