EdamamePod3D: A 3D point cloud dataset of edamame pods for non-destructive scale and quality measurement

Description

EdamamePod3D is a three-dimensional (3D) point cloud dataset of harvested edamame (immature soybean) pods designed to support research in pod-level shape analysis, seed segmentation, and quality-related trait extraction. The dataset comprises 80 pods from a single cultivar (Yuagari-musume), harvested on three dates during the 2025 growing season. For each pod, a 3D mesh model was acquired using a smartphone-based scanning system (iPhone 16 Pro Max with Scaniverse) and converted into a metrically scaled, colored point cloud using CloudCompare. The dataset provides textured 3D mesh models, dense colored point clouds, and tabular reference measurements, including seed moisture content and maximum pod thickness obtained with digital calipers, linked to each point cloud by sample ID. These reference measurements enable validation of dimensional features extracted from the 3D models and support analyses of the relationship between pod shape and maturity-related quality traits at the individual-pod level.

Steps to reproduce

This section describes how the data products provided in this dataset (3D mesh models, point clouds, and the accompanying physical/chemical reference measurements) were generated. The workflow is fully reproducible using the files included in the repository together with standard 3D processing software (CloudCompare) and routine laboratory equipment. Step 1: 3D Model Acquisition Each edamame pod was placed horizontally on a flat platform under standard fluorescent room lighting. A smartphone (iPhone 16 Pro Max, Apple Inc.) running the Scaniverse application (Niantic) in Mesh mode was moved continuously around the sample at a scanning distance of approximately 30 cm, covering a hemispherical range of viewing angles so that no region of the pod surface was missed. Scanning was completed within one minute per sample. The reconstructed 3D mesh was exported in OBJ format (meshes/*.obj). Step 2: Mesh-to-Point-Cloud Conversion Each OBJ mesh was converted into a colored 3D point cloud using the Sample points function of CloudCompare (version 2.13.2). The resulting point clouds were exported in binary PLY format with XYZ coordinates and RGB color information (point_clouds/*.ply). Step 3: Maximum Pod Thickness Measurement After 3D scanning, the maximum thickness of each pod was measured using digital calipers (BDC 150, AS ONE). For each of the two seed-containing regions within a pod, the thickest point was measured once, and the larger of the two values was recorded as the maximum pod thickness for that sample (tabular_data/pod_measurements.csv, column caliper_max_pod_thickness_mm). Step 4: Seed Moisture Content Measurement The two seeds from each pod were placed together in a vacuum-sealed pack and heated in a water bath at approximately 90 °C for 3 min, then cooled in ice water. After removing the seed coats and wiping surface moisture, the seed weight after heat treatment (wb, g) was measured using an analytical balance (GR-60, AS ONE; readability 0.0001 g). Seeds were then frozen in liquid nitrogen and freeze-dried for approximately 3 days using a freeze dryer (FDU-12AS, AS ONE), after which the dry weight (wd, g) was measured. Moisture content (%) was calculated as: moisture content (%) = (wb − wd) / wb × 100 The resulting pod-level moisture content is recorded in measurements/pod_measurements.csv (column moisture_content_percent). Step 5: File Linkage and Integrity Check Each pod was assigned a sample ID (e.g., A-1), used consistently across the mesh file (meshes/*.obj), point cloud file (point_cloud/*.obj), and the sample_id column of pod_measurements.csv.

Institutions

• Niigata University

  Niigata, Niigata

Categories

Funders

• JSPS KAKENHI

  Grant ID: JP25K22399
• JSPS KAKENHI

  Grant ID: JP25K02127

Licence