A Data-Driven Image Extraction and Analysis Pipeline for Plant Phenotyping in Controlled Environments

Fahimeh Orvati Nia^1*, Joshua Peeples^1*, Seth C. Murray², Andrew McFarland³, Troy Vann³, Shima Salehi¹, Robert Hardin⁴, David D. Baltensperger², Amir Ibrahim³, J. Alex Thomasson⁵, Henry Fadamiro⁶, Nithya K Subramanian², Rebecca Roston⁷, Joslin Ishimwe⁷, Diptadeep Basak⁷, Nazar Oladepo¹, and Uday Vysyaraju¹

¹Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX, USA
²Department of Soil and Crop Sciences, Texas A&M University, College Station, TX, USA
³Texas A&M AgriLife Research, Texas A&M University, College Station, TX, USA
⁴Department of Biological and Agricultural Engineering, Texas A&M University, College Station, TX, USA
⁵Department of Agricultural and Biological Engineering, Mississippi State University, Mississippi State, MS, USA
⁶Department of Entomology, Texas A&M University, College Station, TX, USA
⁷Department of Biochemistry, University of Nebraska–Lincoln, Lincoln, NE, USA

^*Correspondence email: fahimehorvatinia@tamu.edu, jpeeples@tamu.edu

Materials

Paper

Dataset

Code

Hugging Face Demo

Documentation

Institution

Abstract

Temporal imaging of plants in controlled environments helps scientists better understand growth and biological processes. However, analyzing large volumes of images has been limited by a lack of standardized datasets and open-source tools. This paper presents an automated multispectral image extraction and analysis pipeline coupled with the Plant Growth and Phenotyping Version 2 (PGP v2) dataset, which provides a comprehensive resource for plant phenotyping research. The pipeline is designed to streamline the workflow from raw greenhouse imaging data to extracted phenotypic traits. The PGP v2 dataset contains approximately 52,000 multispectral images from four crop species (corn, cotton, rice, and sorghum) with labeled metadata. We demonstrate the effectiveness of the proposed methodology through statistical analysis of the extracted features and present comparative analyses of multiple deep learning-based segmentation models. The proposed pipeline and dataset aim to accelerate research in plant phenotyping, facilitate reproducible science, and provide researchers with reliable tools to extract meaningful plant phenotypes from imaging data.

Project Organization and Workflow Integration

**Figure 1:** Integration of teams and responsibilities in the Texas A&M AgriLife Phenotyping Greenhouse project.

Comprehensive Phenotyping Pipeline

**Figure 2:** Comprehensive pipeline for multispectral plant phenotyping and feature analysis.

Plant Growth and Phenotyping Version 2 Dataset

This study introduces the expanded Plant Growth and Phenotyping Version 2 dataset (PGP v2), which substantially increases both scale and diversity from the initial release, comprising approximately 52,000 multispectral images across four crop species. The dataset was collected in the Texas A&M AgriLife Precision Phenotyping Greenhouse and includes temporal sequences with standardized lighting, environmental controls, and imaging protocols. Each image is accompanied by comprehensive metadata including timestamp, plant genotype, growth stage, and environmental conditions. The dataset is divided into training and validation subsets to support reproducible machine learning model development and evaluation.

Dataset Samples

Corn (~14,000 images)

Cotton (~27,000 images)

Rice (~1,376 images)

Sorghum (~10,608 images)

Citation

Plain Text:

F. Orvati Nia, J. Peeples, S. C. Murray, A. McFarland, T. Vann, S. Salehi, R. Hardin, D. D. Baltensperger, A. Ibrahim, J. A. Thomasson, H. Fadamiro, N. K. Subramanian, R. Roston, J. Ishimwe, D. Basak, N. Oladepo, and U. Vysyaraju. 
"A Data-Driven Image Extraction and Analysis Pipeline for Plant Phenotyping in Controlled Environments." 
bioRxiv, 2026.

BibTex:

@article{orvati2026data,
  title={A Data-Driven Image Extraction and Analysis Pipeline for Plant Phenotyping in Controlled Environments},
  author={Orvati Nia, Fahimeh and Peeples, Joshua and Murray, Seth C and McFarland, Andrew and Vann, Troy and Salehi, Shima and Hardin, Robert and Baltensperger, David D and Ibrahim, Amir and Thomasson, J. Alex and Fadamiro, Henry and Subramanian, Nithya K and Roston, Rebecca and Ishimwe, Joslin and Basak, Diptadeep and Oladepo, Nazar and Vysyaraju, Uday},
  journal={bioRxiv},
  pages={2026--02},
  year={2026},
  publisher={Cold Spring Harbor Laboratory}
}