Kurdish Fig Leaf Dataset (KFLD)

Description

Kurdish Fig Leaf Dataset (KFLD) This dataset comprises fig leaf photographs gathered from orchards in Sartak Valley, located near Mount Bamo in the Kurdistan Region of Iraq (roughly 80 km from Sulaimaniyah). The image collection spanned three days, from 8:00 AM to 6:00 PM, to capture a range of natural lighting and shadow patterns. Trees surveyed ranged from 3 to 35 years old, providing diversity in leaf structure and disease progression. Photographs were taken with a Nikon D7100 DSLR camera (6000×4000 resolution, 24-bit RGB color) using consistent settings—ISO 100, f/5, 1/125s shutter speed, 30–75mm focal length, and no flash—to preserve natural lighting conditions. File sizes ranged from 5 to 7MB per image. In total, 2,037 original images were collected, spanning four classes: Mosaic (551 images), Anthracnose (448 images), Cercospora Leaf Spot (525 images), and Healthy leaves (513 images). All photos were taken in real orchard settings, capturing natural differences in background, leaf angle, and disease intensity, then standardized to a uniform resolution. The dataset was further expanded using four augmentation techniques applied independently: horizontal flipping, rotation (±20°), brightness variation (±20%), and contrast variation (±20%). This process generated 8,148 new images, bringing the total dataset size to 10,185 images, all uniformly sized at 512×512 pixels. KFLD is designed to advance AI-driven and computer vision research for automated diagnosis of fig leaf diseases, supporting applications such as classification, object detection, segmentation, disease severity estimation, and model development for precision agriculture.

Steps to reproduce

Photographs were taken with a Nikon D7100 DSLR camera (6000×4000 resolution, 24-bit RGB color) using consistent settings—ISO 100, f/5, 1/125s shutter speed, 30–75mm focal length, and no flash—to preserve natural lighting conditions. File sizes ranged from 5 to 7MB per image. In total, 2,037 original images were collected, spanning four classes: Mosaic (551 images), Anthracnose (448 images), Cercospora Leaf Spot (525 images), and Healthy leaves (513 images). All photos were taken in real orchard settings, capturing natural differences in background, leaf angle, and disease intensity, then standardized to a uniform resolution.

Institutions

• Sulaimani Polytechnic University

  Sulaymaniyah, Sulaymaniyah
• University of Halabja

  Sulaymaniyah, Halabja

Categories

Licence