A Medicinal Plant Leaf Image Dataset of 13 Species Collected in Bangladesh

Description

This dataset contains leaf images of 13 medicinal plant species collected from the National Botanic Garden of Bangladesh, Boldha Garden (Wari, Dhaka, Bangladesh), and Aftabnagar (Dhaka, Bangladesh). Images were acquired between July and November using Apple iPhone 15 Pro Max and Apple iPhone 14 smartphone cameras under natural daylight conditions. To ensure species authenticity, image collection was conducted with guidance from personnel at the National Botanic Garden of Bangladesh and experts associated with the Bangladesh National Herbarium. The dataset comprises 4,063 original images captured from multiple viewing angles and standardized to a resolution of 512 × 512 pixels in JPG format. An augmented version containing 13,000 images is also provided. Augmentation techniques include horizontal flipping, rotation, brightness adjustment, contrast modification, zooming, translation, and Gaussian blurring. The dataset is organized in a class-wise folder structure and includes both original and augmented image collections. It is intended to support research in plant species recognition, computer vision, machine learning, deep learning, transfer learning, self-supervised learning, biodiversity informatics, and precision agriculture. Dataset Contents • 13 medicinal plant species • 4,063 original leaf images • 13,000 augmented leaf images • JPG image format • Resolution: 512 × 512 pixels • Class-wise folder organization Keywords: medicinal plants, leaf images, plant species recognition, computer vision, image classification, deep learning, biodiversity informatics, Bangladesh.

Steps to reproduce

1. Visit the National Botanic Garden of Bangladesh and identify medicinal plant species with assistance from botanical experts or reference herbarium records. 2. Collect representative leaf specimens from the selected medicinal plant species. 3. Place each leaf on a white background to minimize background interference during image acquisition. 4. Capture images using smartphone cameras (Apple iPhone 15 Pro Max and Apple iPhone 14) under natural daylight conditions during morning and afternoon hours. 5. Acquire multiple images from different viewing angles to capture variations in leaf morphology, texture, venation patterns, and color characteristics. 6. Manually inspect all collected images and remove blurred, duplicate, damaged, or low-quality samples. 7. Crop and resize the retained images to a uniform resolution of 512 × 512 pixels. 8. Save all processed images in JPG format and organize them into class-specific folders corresponding to the medicinal plant species. 9. Apply offline data augmentation techniques, including horizontal flipping, rotation, brightness adjustment, contrast modification, zooming, translation, and Gaussian blurring. 10. Generate approximately 1,000 images per class to obtain the augmented dataset containing 13,000 images. 11. Store both the original and augmented datasets in a hierarchical folder structure for subsequent computer vision, machine learning, and deep learning applications.

Institutions

• East West University

  Dhaka Division, Dhaka

Categories

Licence