BPLD: Bangladeshi Plant Leave Dataset

Overview The Bangladeshi Plant Leave Dataset is a self-collected hierarchical plant leaf image dataset designed to support research in hierarchical image classification, plant species recognition, and deep learning-based agricultural applications. The dataset contains 7,200 leaf images organized into a two-level hierarchical taxonomy consisting of three primary classes Flower, Fruit, and Medicinal plants and eighteen fine-grained subclasses. All images were captured in real-world outdoor environments using a personal mobile phone camera, ensuring natural variations in illumination, orientation, background complexity, and scale. The dataset is balanced at both hierarchical levels, making it suitable for multi-stage classification frameworks and comparative deep learning studies. Dataset Structure The dataset follows a two-stage hierarchical structure: Level 1 – Main Classes (3 classes) : Flower Fruit Medicinal Level 2 – Subclasses (6 per main class): Flower: Hibiscus, Lotus, Marigold, Rose, Rose Periwinkle, Togor. Fruit: Guava, Jackfruit, Lemon, Mango, Papaya, Wood Apple. Medicine: Bitter Vine, Neem, Pathorkuci, Thankuni, Tulshi, Vasaka. Dataset Statistics Total Images: 7,200 Total Main Classes: 3 Total Subclasses: 18 Images per Subclass: 400 Images per Main Class: 2,400 Class Distribution: Fully balanced Image Acquisition Protocol All images were captured by the dataset creator using a personal mobile device under natural environmental conditions. The acquisition process intentionally preserved real-world variability, including differences in lighting, background clutter, leaf pose, and scale. No synthetic image generation was used. This real-world collection strategy enhances the dataset’s applicability for robust deep learning model development. Intended Use The dataset is intended for: Hierarchical image classification research Multi-stage deep learning frameworks Transfer learning evaluation Fine-grained plant leaf recognition Agricultural and medicinal plant identification systems