FloraGuard: a CNN-based fungal detection system for smart horticulture

Fungal diseases represent one of the most damaging threats to ornamental plant productivity and economic value, particularly in nursery and horticultural sectors. Traditional detection methods rely heavily on manual inspection, which is inefficient, inconsistent, and inaccessible to non-specialists. This project introduces FloraGuard, a novel lightweight detection system powered by Convolutional Neural Network (CNN) architecture using MobileNet-V2, designed to automate the identification of four common fungal diseases powdery mildew, black spot, rust, and botrytis blight from leaf images. A curated dataset of 568 labelled images was preprocessed through normalization, augmentation, and filtering to enhance model robustness. The final trained model demonstrated 95.73% accuracy, with high F1-scores across all classes, and was integrated into a user-friendly GUI accessible via web or mobile interface. FloraGuard has potential to reduce fungicide overuse, minimise losses, and democratise plant disease management for small-scale growers. Its relevance aligns with SDG Goals 2 and 12 and provides a scalable foundation for commercial and academic extensions.