GrainShape: A landmark-annotated image dataset of rice grains for geometric morphometric analysis.

This open-access dataset comprises matched image and landmark files for 231 rice grains representing distinct landraces cultivated in China's Lake Taihu region. Grains were grown in a single 2022 field season under uniform management. One mature, fully-filled kernel per accession was photographed dorsoventral side up with an Olympus TG-6 macro camera (4000 × 3000 px, 314 dpi, f/2.3, 1/13 s, +1 EV) under fixed laboratory lighting. Each frame includes a matte-green background, a 30 mm plastic ruler, and two printed fiducials, enabling pixel-to-millimeter scaling and consistent orientation. The repository is organized into four zip archives. grainimages.zip holds the 231 color JPEG photographs. landmarks.zip provides 231 plain-text tables listing the x,y coordinates of 30 homologous radial landmarks that capture length, width, curvature, tip geometry, shoulder position, awn insertion point, and subtle asymmetries. contours.zip contains 231 standalone PNG drawings showing a spline-interpolated outline reconstructed solely from each landmark set-no photographic background is included-facilitating rapid visual checks of digitization accuracy. scripts.zip contains all custom Python scripts for generating the landmark sets and reproducing the figures and tables presented in the paper. The dataset, available on Zenodo (DOI:10.5281/zenodo.15515366) under a CC BY 4.0 license, can be reused for landmark- or outline-based morphometric analysis, color and awn-length phenotyping, benchmarking of computer-vision pipelines, and instructional demonstrations of geometric morphometric workflows.