Analyzing and Classifying Mango Leaf Diseases Using Advanced Image Processing and Deep Learning Techniques

The performance of several deep learning models for classification tasks is thoroughly assessed in this paper, with an emphasis on contrasting conventional architectures with a suggested hybrid model. DenseNet121, VGG19, ResNet50, InceptionV3, AlexNet, MobileNetV2, and EfficientNetV are among the models that were evaluated using important metrics like accuracy, precision, recall, and F1-score. The most balanced and successful model overall was DenseNet121, which outperformed the baseline models with the maximum accuracy of 95.33% and a perfect recall of 1.00. Additionally, ResNet50 and VGG19 performed well, each obtaining an F1-score of 0.95. With a precision score of 0.96, the suggested hybrid model was the most notable, demonstrating its potent capacity to lower false positives. Nevertheless, its F1-score (0.91) was marginally affected by its somewhat poorer recall (0.88). Even yet, the hybrid model showed that mixing different architectures can result in improved performance and exceeded some standard models in terms of overall accuracy (91.66%). The findings demonstrate that although hybrid approaches can capitalize on complementary assets to attain competitive outcomes, traditional models continue to be formidable competitors. For deep learning model selection and design, this comparison provides insightful information, particularly for applications that need high precision and balanced performance across evaluation measures.