Automated Severity Assessment of Algal Leaf Spot in Guava using CNN and Logistic Regression

This research presents a novel approach to detect and classify different threat levels of guava leaf diseases using Convolutional Neural Networks (CNNs). The proposed CNN model, comprising three convolutional layers, max-pooling layers, and fully connected layers, effectively extracts and processes relevant features from the image data. The model was trained and evaluated on a dataset of $\mathbf{1 0, 2 7 0}$ images, achieving impressive performance metrics. The precision, recall, and F1score for each threat level exceeded 94%, indicating high accuracy and robustness. The model’s ability to handle class imbalances was demonstrated by the consistent performance across different evaluation metrics, including macro, weighted, and micro averages. This research has significant implications for the agricultural industry, particularly in the cultivation of guava. By accurately diagnosing leaf diseases, farmers can implement timely and effective control measures, leading to improved crop yield and quality.