3D Point Cloud Dataset of Faba Bean (Vicia faba L.) with SPAD and Biomass Ground Truth

A 3D photogrammetry dataset of five faba bean genotypes, captured from handheld camera scans and reconstructed into segmented point clouds. Each plant is paired with manually measured ground-truth values for leaf chlorophyll (SPAD) and above-ground shoot dry biomass. The dataset accompanies the paper "Non-destructive Biomass and SPAD Estimation Based on 3D Photogrammetry for Faba Bean" (EE&AE 2025) and is intended for plant phenotyping, point-cloud regression, and deep-learning research. Dataset overview Property Value Crop Faba bean (Vicia faba L.) Genotypes Genius, Fanfare, LYNX, Ghengis, Vertigo (5) Number of plants 36 Data type Segmented 3D point clouds (.ply) + tabular ground truth (.xlsx) Point attributes XYZ coordinates + RGB colour Measured traits SPAD (chlorophyll), shoot dry biomass (g) Acquisition Handheld camera video → NeRF reconstruction → crop segmentation Location University of Lincoln, Riseholme campus polytunnel, UK Year 2024 Files and structure . ├── README.md # This file ├── ground_truth.xlsx # SPAD and biomass measurements for all 36 plants └── Only Plants/ # 36 segmented point cloud files ├── Fanfare_1.ply ├── Fanfare_2.ply ├── ... ├── Genius_1.ply ├── Ghengis_1.ply ├── LYNX_1.ply └── Vertigo_7.ply Point cloud files (.ply) Each *.ply file is a single segmented faba bean plant. The pot, stool, stake, label, scale object and soil have been removed, leaving only the plant. Each point stores: x, y, z — spatial coordinates (metric scale) red, green, blue — RGB colour (0–255) File names follow the convention <Genotype>_<PlantNumber>.ply, which matches the Genotype column in ground_truth.xlsx exactly (e.g. Ghengis_4.ply ↔ row Ghengis_4). Ground truth (ground_truth.xlsx) A single sheet with 36 rows and the following columns: Column Description No Sample index (1–36) Genotype Plant identifier, matches the .ply file name SPAD Leaf chlorophyll meter reading (SPAD units) Shoot dry biomass(g) Above-ground dry biomass in grams Per-genotype summary Genotype Plants SPAD (mean ± SD) Biomass g (mean ± SD) Fanfare 6 46.93 ± 2.94 6.76 ± 3.01 Genius 5 49.40 ± 4.21 6.70 ± 1.93 Ghengis 9 49.74 ± 4.50 8.11 ± 3.36 LYNX 9 48.52 ± 4.29 4.30 ± 1.16 Vertigo 7 50.29 ± 1.93 8.70 ± 2.93 All 36 SPAD range 41.9–57.6 Biomass range 2.15–12.90 Data acquisition Seeds were stratified in Petri dishes until radicle emergence, inoculated to enhance root nodulation, and sown on 1 July 2024 in 4-liter pots inside a polytunnel at the University of Lincoln's Riseholme campus. Plants were regularly irrigated, with weed, disease, and pest control applied throughout. Plants were digitally scanned with a handheld camera on 23 September 2024. The 3D point clouds were reconstructed from the video streams using a Neural Radiance Field (NeRF) pipeline, then post-processed with a color- and geometry-aware 3D segmentation method followed by manual refinement to remove all non-plant elements. Reference measurements were taken on the same plants: SPAD with a digital chlorophyll meter, and shoot dry biomass after oven-drying at 62 °C for 48 hours. Loading example (Python) import open3d as o3d import pandas as pd gt = pd.read_excel("ground_truth.xlsx") pcd = o3d.io.read_point_cloud("Only Plants/Ghengis_4.ply") # ground truth for this plant row = gt[gt["Genotype"] == "Ghengis_4"].iloc[0] print(row["SPAD"], row["Shoot dry biomass(g)"]) Citation If you use this dataset, please cite the accompanying paper: M. Kaya, L. Guevara, J. Singh, G. Cielniak, A. Yilmaz, and R. Valluru, "Non-destructive Biomass and SPAD Estimation Based on 3D Photogrammetry for Faba Bean," 2025 10th International Conference on Energy Efficiency and Agricultural Engineering (EE&AE), Starozagorski Bani, Bulgaria, 2025. DOI: 10.1109/EEAE65901.2025.11273533 And cite this dataset via its Zenodo DOI: 10.5281/zenodo.20788756 Funding This work was partly supported by the Agri-OpenCore (grant ID: 10041179) and NUE-Profits (grant ID: 10044243) projects. Contact School of Agri-Food Technology and Manufacturing, University of Lincoln, UK. Contact: Murat Kaya — [e-mail] or [e-mail]