Automatic Detection of Leaf Diseases in Hibiscus Plants Using Live Image Dataset with User Interface

The agriculture sector globally faces significant challenges due to plant diseases, with ensuring crop loss and economic barriers. Hibiscus plants, grown as both ornamental and medicinal plants, are affected by diseases that can damage the various parts of the plant and reduce their growth and yield. In this research, we suggest an approach to automatically identify hibiscus plant diseases, using deep-learning strategies. Our proposed work involves the development of a Convolutional Neural Network model trained on an extensive dataset that included images of hibiscus leaves infected with plant diseases, as well as images of healthy hibiscus leaves. Our model leveraged Transfer learning and it showed promise in the detection of the disease and the type of disease in the hibiscus leaves. The significant advantage of this approach is the early detection of hibiscus plant diseases, which is an opportunity for hibiscus disease management and loss reduction and improved health of the plant. The novelty of our approach lies in its focus on the distinct visual markers of hibiscus plant diseases, such as discoloration, shrinkage, and fungal spores like Anthracnose and leaf rust, which are critical in determining plant health. Also, it facilitates early detection, enabling timely intervention with minimal pesticide use. This ensures healthier plantations and contributes to sustainable cultivation practices. Furthermore, the insights gained from this study can serve as a blueprint for addressing plant diseases in other agricultural species, thereby advancing global food security. Our model achieved a high test accuracy of 97.21%, with a precision of 96.88%, recall of 96.45%, and an F1 score of 96.66%. Additionally, the model demonstrated robust performance in distinguishing between healthy and diseased leaves, as evidenced by a ROC-AUC score of 98.50%, underscoring its effectiveness in disease detection for hibiscus plants.