Leaf Image Dataset for Plant disease Classification and Agricultural Analysis

Description

This dataset contains 1,545 leaf images collected from 7 different crops, including both healthy and diseased leaves. The images were captured in different environmental conditions to make the dataset more diverse and useful for real-world applications. The dataset contains images of Bitter gourd (303)- Bunchy top of bitter gourd (12), Downy mildew (118), Leaf curl(mosaic virus) (132), Healthy leaf (41); Brinjal (298)- Cercospora leaf spot (110), Early blight (81), Leaf curl (58), Healthy leaf (49); Ladies finger (252)- Cercospora leaf spot (41), , Leaf curl(68), mosaic virus (46), Healthy leaf (44); Mung bean (135)- caterpillars (Phaseolus vulgaris) (37), Mung bean Yellow Mosaic Virus (MYMV) (12), Leaf curl (50), Healthy leaf) (36); Sesame (222)- Bacterial leaf spot (Pseudomonas sesami) (36), caterpillars(Phaseolus vulgaris)(65), Sesame yellow mosaic virus (SYMV)(27), Leaf curl(38), Healthy leaf(56); Snake gourd(153)- (Anthracnose (Colletotrichum spp.) (31), Leaf Curl(54) Yellow Mosaic Virus (35), Healthy leaf (33)and Yard long bean-(182) (Cercospora leaf spot (55), leaf curl virus (48), yellow mosaic virus (48), Healthy leaf (31).All images were resized to 256×256 pixels for faster and more efficient processing. After preprocessing, the dataset was augmented and expanded to a total of 12,360 images while maintaining class balance. This dataset can be used for image classification, transfer learning, and deep learning models such as CNNs and transformers in plant disease detection and precision agriculture research.

Steps to reproduce

The Multi-Class Plant Disease Dataset (Seven Classes) was collected from agricultural fields by capturing infected leaf images under natural daylight conditions using a OnePlus 3 (A3003) smartphone. Leaves with clear disease symptoms were selected, and disease identification was verified by Associate Professor Dr. Ziaur Rahman Bhuiyan, Department of Plant Pathology, and Kbd. MD. Jahidul Islam, Research Officer, Institute of Seed Technology, Sher-e-Bangla Agricultural University. Most samples were photographed on a plain white background to reduce noise and improve image clarity. The dataset was manually inspected, annotated, and organized into labeled folders for each disease class. Blurred, low-quality, and incorrectly labeled images were removed, and image resolutions were standardized while preserving visual quality. The final dataset provides a structured and high-quality resource suitable for disease classification, feature extraction, transfer learning, and AI-based agricultural research, with a reproducible data collection and preprocessing methodology for future extensions.

Institutions

• Stamford University Bangladesh

  Dhaka Division, Dhaka

Categories

Licence