Sunflower Leaf Disease Image Dataset for Automated Plant Pathology

The Sunflower Leaf Disease Image Dataset for Automated Plant Pathology contains high-resolution images of sunflower leaves collected from agricultural fields in Bangladesh, designed for detecting and classifying multiple sunflower leaf diseases. Images were captured in-field using a Redmi Note 11 smartphone under natural daylight and varied weather conditions to ensure diversity in lighting and environment. The dataset supports research in deep learning, computer vision, and plant pathology, including healthy leaves and four disease classes suitable for multi-class classification.

The original dataset includes 1,053 images distributed as follows: Alternaria Leaf Spot (203), Downy Mildew (217), Healthy (194), Powdery Mildew (237), and Wilted Leaves (202). Each image has an ultra-high resolution of 12,288 × 16,320 pixels (\~25 MB), totaling approximately 14.2 GB. Due to the large size, all images were resized to 480 × 637 pixels for practical use.

Data were collected at Daffodil Smart City and Model Town Nursery in Birulia, Ashulia, Savar, Dhaka, Bangladesh, using a Redmi Note 11 camera with 24-bit color depth, aperture f/1.8, stored in JPEG format.

For data augmentation, OpenCV techniques such as rotation (±15°, ±30°), flips, zoom scaling (1.2×, 1.5×, 1.8×), translation (±10 px), cropping (0.6, 0.8), and various blurs were applied to expand the dataset, resulting in approximately 500 images per class and a total of 2,500 images. Example metadata for an image (IMG\_20250522\_174724.jpg) includes capture with a Xiaomi camera model 23117RA86G, 24-bit sRGB color, exposure 1/100 sec, ISO 200, focal length 6 mm, auto white balance, GPS latitude 23.5247046, longitude 90.1918097, and altitude 34.5 m. The JPEG file size is 26.1 MB.

The dataset has been validated by a Sub-Assistant Agriculture Officer from the Department of Agricultural Extension (DAE), Bangladesh.