Aerial Imagery and Segment Anything Model for Architectural Trait Phenotyping to Support Genetic Analysis in Peanut Breeding

<p>Multimodal UAV imagery datasets for ML-based classification of growth habit and mainstem prominence in peanut, associated with the manuscript “Aerial Imagery and Segment Anything Model for Architectural Trait Phenotyping to Support Genetic Analysis in Peanut Breeding,” which has been accepted for publication and is available online in <em>Plant Phenomics</em> as of 27 October 2025 (<a href="https://doi.org/10.1016/j.plaphe.2025.100126" rel="noopener" target="_blank">https://doi.org/10.1016/j.plaphe.2025.1001266</a>). These datasets provide plot-level images and categorical labels for peanut architectural traits used in ML classification experiments.</p> <p>Data were collected from peanut breeding plots at the University of Georgia during the 2022 growing season. RGB images were acquired using UAV flights. Normalized Digital Surface Models (nDSM), representing the height of plants above ground, were derived from the RGB imagery using photogrammetric reconstruction. Pseudo-RGB images were generated by embedding the nDSM information into the RGB images, combining color and structural information.</p> <p>Images are organized into training (80%) and testing (20%) subsets. Trait labels are encoded in folder names, and no genotype information is included.</p> <p>Growth Habit (GH) classes:</p> <ul> <li> <p>Training: <em>Bunch</em>, <em>Mixed</em>, <em>Spreading</em>, <em>Spreading and Bunch</em>, <em>Prostrate*</em></p> </li> <li> <p>Testing: <em>Bunch</em>, <em>Mixed</em>, <em>Spreading</em>, <em>Spreading and Bunch</em></p> </li> </ul> <p>Mainstem Prominence (MP) classes:</p> <ul> <li> <p><em>Apparent</em>, <em>Non-apparent</em>, <em>Somewhat Apparent</em></p> </li> </ul> <p>*Note: The <em>Prostate</em> class is only present in the GH training set and is not included in the test set.</p> <h3><strong>Associated Manuscript</strong></h3> <p>Rodriguez-Sanchez, J., Da Silva, R. M., Chu, Y., Rodriguez, L., Zhang, J., Johnsen, K., Ozias-Akins, P., & Li, C. (2025). Aerial Imagery and Segment Anything Model for Architectural Trait Phenotyping to Support Genetic Analysis in Peanut Breeding. Plant Phenomics, 100126. <a href="https://doi.org/10.1016/j.plaphe.2025.100126" rel="noopener" target="_blank">https://doi.org/10.1016/j.plaphe.2025.1001266</a></p> <h3><strong>Authors</strong></h3> <p>Javier Rodriguez-Sanchez, Raissa Martins Da Silva, Ye Chu, Lenin Rodriguez, Jing Zhang, Kyle Johnsen, Peggy Ozias-Akins, and Changying Li<strong><br></strong></p> <h3><strong>License</strong></h3> <p>Creative Commons Attribution 4.0 International (CC BY 4.0)</p> <h3>Intended Use</h3> <p>This dataset supports machine learning classification of peanut architectural traits and is suitable for research in high-throughput phenotyping, multimodal image analysis, and canopy architectural trait quantification. It can be used for training, evaluation, and benchmarking of supervised and multimodal deep learning models.</p> <h3>Folder Structure</h3> <p>Dataset/<br>├── Growth-habit-dataset/<br>│   ├── RGB/<br>│   │   ├── train/<br>│   │   │   ├── Bunch/<br>│   │   │   ├── Mixed/<br>│   │   │   ├── Spreading/<br>│   │   │   ├── Spreading-and-Bunch/<br>│   │   │   └── Prostrate/   # Only in training set<br>│   │   ├── test/<br>│   │   │   ├── Bunch/<br>│   │   │   ├── Mixed/<br>│   │   │   ├── Spreading/<br>│   │   │   └── Spreading-and-Bunch/<br>│   ├── nDSM/            # Same train/test structure as RGB<br>│   └── pseudoRGB/       # Same train/test structure as RGB<br>│<br>├── Mainstem-prominence-dataset/<br>│   ├── RGB/<br>│   │   ├── train/<br>│   │   │   ├── Apparent/<br>│   │   │   ├── Non-apparent/<br>│   │   │   └── Somewhat-apparent/<br>│   │   ├── test/<br>│   │   │   ├── Apparent/<br>│   │   │   ├── Not-apparent/<br>│   │   │   └── Somewhat-apparent/<br>│   ├── nDSM/            # Same train/test structure as RGB<br>│   └── pseudoRGB/       # Same train/test structure as RGB<br>│<br>└── README.txt</p>