A Fine-Tuned MobileNetV2 Deep Learning Model for Citrus Fruit Disease Classification

In this article, they describe a unique strategy for recognizing orange disorders leveraging the MobileNetV2 model, a lightweight convolutional neural network architecture. Our technique covers thorough measures to assure robustness and accuracy in illness categorization. They standardized picture quality and format, implemented data augmentation methods, and stratified the dataset into training, validation, and test sets. Leveraging MobileNetV2's efficiency and adaptability for mobile applications, it initialized the model with pre-trained weights from ImageNet and altered the classification layer to incorporate illness classifications. Through transfer learning, they fine-tuned the model on our orange disease dataset, applying a learning rate scheduler and stochastic gradient descent optimizer with momentum to optimize training. During training, they tracked progress using measures like accuracy, loss, precision, recall, and F1-score on the validation set. Early halting procedures were added to avoid overfitting and guarantee model generalization. Our trained model attained a validation accuracy of 98.66% and a validation loss of 0.1145, proving its usefulness in properly diagnosing orange disorders. Furthermore, they tested model performance on the test set, computing additional performance measures like accuracy, recall, and F1-score to quantify its effectiveness. Visualization tools such as confusion matrices were applied to acquire insights into model predictions and possible areas for development. last packaged the trained MobileNetV2 model for deployment in real-world applications, combining it with user-friendly interfaces to allow smooth interaction for growers and agricultural practitioners. Thorough testing and validation methods were done to assure the dependability and robustness of our solution in actual field circumstances. Overall, our work demonstrates the potential of deep learning approaches, notably MobileNetV2, in revolutionizing disease diagnosis in agriculture, consequently increasing crop management practices and encouraging sustainable agricultural growth.