A Medicinal Plant Leaf Image Dataset from Bangladesh for Deep Learning–Based Recognition
Description
This dataset contains a total of 4,741 JPG images of medicinal plant leaves collected from the rural jungle area of Guadanga village, Phulpur, Mymensingh, Bangladesh. The images were captured in indoor conditions under controlled lighting using two high-resolution smartphone cameras - Google Pixel 5 and Google Pixel 6a - to ensure high-quality visuals suitable for computer vision and deep learning applications. The original dataset is organized into six folders, each representing a distinct and commonly found medicinal plant species in Bangladesh: • Centella Asiatica • Coccinia Grandis • Eclipta Prostrata • Mikania Micrantha • Murraya Koenigii • Stephania Japonica
Files
Steps to reproduce
A total of 1,678 original JPG images were captured and supplemented with 3,063 augmented images generated using standard image augmentation techniques, including flipping, grayscale conversion, and geometric transformations. These augmentations improve data diversity and support the training of more robust deep learning models for leaf classification and recognition tasks. Object Detection and Annotation: Object detection was performed using Roboflow, where leaf regions were annotated using bounding boxes to facilitate tasks such as detection, segmentation, and recognition. The annotations were manually verified to maintain consistency and labeling accuracy across all six plant classes. Dataset Split: The dataset is divided into three subsets to support model training and evaluation: • Training Set: 2,363 images (77%) • Validation Set: 400 images (13%) • Test Set: 300 images (10%) This stratified split ensures balanced representation of all plant classes across each subset. Preprocessing and Augmentation Details: • Auto-Orientation: Applied to all images • Resize: Stretched uniformly to 640×640 pixels • Augmentation Outputs per Example: 3 • Applied Augmentations: o Horizontal and vertical flips o Grayscale applied to 15% of images All images remain in JPG format and retain realistic visual properties, making this dataset highly suitable for use in deep learning–based leaf recognition, medicinal plant classification, image segmentation, and object detection tasks. Attribute Details Total Images 4,741 JPG images Original Images 1,678 Augmented Images 3,063 Plant Species/Classes 6 Capture Devices Google Pixel 5, Google Pixel 6a Image Resolution 640×640 (after preprocessing) Source Location Jungle of Guadanga, Phulpur, Mymensingh, Bangladesh Annotation Tool Roboflow Applications Deep learning, medicinal plant recognition, AI-based botany, image segmentation
Institutions
- American International University Bangladesh
- North South University
- Chittagong University of Engineering and Technology
- Sher-e-Bangla Agricultural University