ClusterBean-FD: a dataset for detection and classification of foliar disease in cluster bean plants
Аннотация
Cluster bean (Cyamopsis tetragonoloba) is a globally significant crop, known for its culinary, industrial, and health benefits. Early and precise identification of foliar diseases in cluster beans is critical for reducing crop losses. Recently, deep learning techniques have revolutionized crop disease detection, offering more efficient and accurate alternatives to traditional methods. Despite the availability of various public datasets for many other crops, there is a scarcity of specialized datasets for cluster bean crop. To address this gap, the presented study introduces a new dataset, named ClusterBean-FD , comprising 9,249 real-world field-condition images of cluster bean plants across 13 foliar disease classes from the Gwalior-Chambal region of Madhya Pradesh, India. This is introduced as the first publicly available dataset specifically designed for the detection and classification of foliar diseases in cluster bean, a crop of considerable industrial and nutritional importance due to its role in guar gum production. To enhance the robustness of disease detection models, the dataset includes images capturing various environmental conditions, such as different lighting, angles, and stages of leaf damage. This dataset uniquely contributes to the field and paves the way for more accurate and reliable disease prediction models. To assess the efficacy and applicability of the curated dataset, meta-analyses for multiple deep learning models based on CNNs architecture were performed. The analysis demonstrated that, among the evaluated CNN models, EfficientNet-B0 achieved the best performance due to its efficient multi-scale feature extraction and a good trade-off between computation and accuracy.