Xception for Tomato Leaf Disease Detection: Hyperparameter Tuning and Fine-tuning Approaches

Tomato cultivation is essential to world agriculture, affecting food security and economic stability. Tomasto leaf diseases must be detected accurately to maintain crop health and maximize yield. This work improves disease classification by using deep learning models, notably CNN, Xception, DenseNet, and a hyper-tuned Xception model. We used transfer learning and data augmentation to improve model performance on the Kaggle dataset, which contains 10,000 training images and 1,000 validation images from the Plant Village Repository organized into ten groups. The hyper-tuned Xception model achieved the maximum performance by meticulously tweaking hyperparameters such as learning rate, batch size, and network architecture, with an accuracy of 97.23%, precision of 95.73%, recall of 94.40%, and F1-score of 95.06%. These results show a considerable improvement over the CNN, Xception, and DenseNet models. The results highlight the efficiency of sophisticated deep learning algorithms and hyperparameter tweaking in achieving high classification accuracy, making the proposed model a robust tool for practical tomato leaf disease diagnostics.