BLADE: Banana Leaf Agricultural Disease Evaluation
Description
This dataset is a detailed table of visual, textural, and spatial features taken from standardized 512x512 RGB images of both healthy and diseased banana leaves collected from the UP-Bihar border region. It is designed to help researchers use machine learning and artificial intelligence in precision agriculture and environmental analysis. Each sample has 31 different morphological and photometric features, which help identify important leaf diseases like Sigatoka and Cordona. To make automated disease classification and severity estimation easier and faster, the dataset includes metadata about basic image properties, such as file size and resolution, as well as detailed color statistics for both RGB and HSV color spaces. This helps algorithms handle changes in lighting during fieldwork. The dataset also contains spatial and intensity measurements, including Gray Level Co-occurrence Matrix (GLCM) statistics like contrast, dissimilarity, homogeneity, energy, and correlation, along with Local Binary Pattern (LBP) features that capture small texture changes and damage typical of necrotic leaf spots. In addition to classifying leaves as healthy or diseased, the dataset allows for accurate grading of disease severity using segmentation metrics like edge pixel counts, spot counts, total spot area, total leaf area, and the percentage of the leaf that is infected. These measures give a clear picture of how the disease progresses.