BellCrop – A Bell Pepper Leaf Dataset for Disease Classification and Yield Enhancement using Machine Learning

The majority of crop vision datasets include domain-specific annotations for use in advanced deep learning applications in both offline and online agricultural settings. In this study, we present BellCrop, a novel dataset of 4,860 high-resolution images of bell pepper leaves divided into three categories: Healthy, Powdery mildew-infected, and Magnesium-deficient leaves. Unlike previous datasets, which were primarily collected in controlled environments, these images were captured in greenhouses under natural conditions. Images were taken with a camera with a resolution of 3468 × 4624 pixels. Deep learning-based classifiers such as VGG19, Xception, and ResNet50 were used to detect these diseases during the early stages of plant growth. Despite the challenging nature of the dataset, the results show detection accuracies of 87%, 81%, and 80%. BellCrop study demonstrates the potential of deep learning models in automating the early detection of leaf diseases in bell peppers. It has significant implications for precision agriculture, allowing for more timely interventions and improved crop management efficiency. Additionally, the study looks into the use of advanced models known as Transformers. EfficientNet-B7 and Vision Transformers (DeiT) achieved more than 90% accuracy rates, improving the reliability of disease classification in agricultural settings.