Daffodil Leaf Disease Recognition Framework: A Unified Approach with CNN and Random Forest Integration

It evaluates the efficacy of an approach to classification across multiple plant disease classes. Each row represents a distinct class, which includes Leaf Scorch, Botrytis Leaf Destruction, Rust, Leaves Smut, Basal Rot, the Penicillium Leaves Blight, Stagonospora Leaves Spot, Xanthomonas Left Spot, and It Mosaics Virus. The evaluation measures include accuracy Recall, F1-Score, Assistance, Support Proportion, & Accuracy. Precision ratings can vary from 91.91% to 94.73%, showing the model's capacity to properly identify occurrences of all categories among positive predictions. Recall scores range from 91.26% to 93.77%, indicating the model's ability to accurately identify all relevant occurrences of each class within the real positive examples. The F1-Score, which is a harmonic mean for Precision and Recall, spans from 92.04% to 94.55%, offering a fair assessment of the algorithm's overall performance in each class. Representing the category distribution of the dataset, the Support column displays the number of examples for each class, ranging from 985 to 1060. The percentage of occurrences for every class over the entire dataset is shown by the Support Proportion. All classes show good overall model performance, as evidenced by the constant 93.14% Weighted Average of Accuracy and Recall with F1-Score. Analyzing the average without taking into consideration differences in class, the Macroeconomic Average comes out at 93.13%. Ninety-one percent is the Micro Average, which gauges overall precision. Plant disease detection is a strong suit for the algorithm, as seen by its reported overall accuracy of 81.83%. The F1-Score, accuracy, and recall statistics demonstrate the model's stable performance and confirm its status as a reliable tool for plant pathology disease classification.