Innovative Fusion for Sunflower Leaf Disease Identification: CNN and Random Forest Strategies

An organizational paradigm to classifying plant diseases into ten categories is examined in detail in this presentation. Downy mildew, rust, mild mildew, Alternaria, septoria, sclerotinia, bacterial, verticillium, fusarium, and cercosporin leaf spots are just a few of the infection categories where the model excels, with recall, precision, and F1-Score evaluates ranging from 91.06% to 94.09%. The accuracy of the model in recognizing various plant diseases is shown by its macro-average of 92.19%, which covers all classes. With a weighted average metric of 92.21% for accuracy, 92.20% for recall, and 92.20% for F1-score, we can conclude that the model is reliable. The model proves its longevity in making correct predictions over the whole dataset with a remarkable 92.20% overall accuracy micro-average. The support percentage and the mean accuracy of 84.1064% show that the model can detect plant diseases even with the dataset's unpredictable occurrences. As a result of this study, we know that the model is accurate and reliable for a variety of diseases, which increases its usefulness for diagnosing plant diseases. The model design allows for the effective identification of plant ailments via the combination of four maximally pooling layers for geographical down-sampling, a single fully linked layer for classification, and a total of four levels of convolution neural network architecture for feature extraction.