The Frangipani–Hibiscus–Marigold–Rose (FHMR) Leaf Image Dataset is a curated, field-collected image resource designed to support reproducible research in plant species identification and computer vision–based agricultural applications. The dataset comprises 1,080 original leaf images (270 per species) captured from healthy plants in nursery environments, along with 6,480 systematically generated augmented images (six per original), resulting in a total of 7,560 images. All images are standardized to 512 × 512 pixels in JPEG format. Image acquisition was conducted in Savar, Dhaka, Bangladesh, using an iPhone 14 Pro camera under both neutral and natural background conditions. Augmentation was performed using a deterministic, parameter-bounded pipeline (random seed = 42), incorporating geometric and photometric transformations to enhance variability while maintaining biological plausibility. In addition to the image files, the dataset repository provides a complete metadata manifest, augmentation scripts, and environment specifications to ensure transparency and repeatability. The FHMR dataset is publicly available via Mendeley Data (DOI: 10.17632/tkgbrvb2c2.4) under a CC BY 4.0 license and serves as a benchmark resource for machine learning–based plant species recognition and related computer vision tasks.
Matin et al. (2026) studied this question.