Bean Leaf Disease Dataset for Multi-Class Image Classification

This dataset contains 1,200 field-collected bean leaf images captured from agricultural plots in Bangladesh for the purpose of multi-class plant disease classification.

The images were collected under real-world environmental conditions, including natural lighting, varying backgrounds, and diverse leaf orientations, ensuring realistic variability for computer vision and deep learning applications.

Data collection was conducted in Sholakura village, Doulotpur union, Belkuchi upazila, Sirajganj district, Bangladesh, on 17th and 18th October 2025. The dataset reflects authentic agricultural conditions and was gathered under the supervision of local agricultural authorities.

The dataset consists of four balanced classes:
- Healthy Leaf (300 images)
- Leaf Spot Disease (300 images)
- Pest Damage (300 images)
- Yellow Mosaic Disease (300 images)

All images are organized into class-specific directories and follow a standardized naming convention. A metadata.csv file is provided, containing labels and predefined train, validation, and test splits (70%, 15%, 15%) using stratified sampling with a fixed random seed to ensure reproducibility.

This dataset is intended for research and academic use in areas such as deep learning, computer vision, plant disease detection, and explainable AI.

The dataset has been validated by agricultural officers and collected under academic supervision, ensuring data authenticity and research reliability.