Classification of Diseases in Mango Leaves based on deep learning using CNN model

Mango leaf diseases cause great harm to mango production in productivity and quality. It is crucial for early and accurate detection of disease to facilitate management and prevention. In this paper, we utilize EfficientNet-B7, MobileNet, and ResNet-152 to introduce a deep learning-based method for automatically classifying mango leaf diseases. The dataset is made up of 60% train, 20% test, and 20% val. For model generalization and prevention of overfitting, extensive data augmentation methods are used in combination with L1_L2 regularization on fully connected layers. The models are learned at the learning rate 0.0001 and with early stopping utilizing the Adam optimizer to avoid overfitting. Measurement of performance for accuracy, confusion matrix, and classification reports display the efficiency with which the resulting models perform. As per experiment results, Utilizing the test dataset, EfficientNet-B7 gives an accuracy of 98.87%, MobileNet has an accuracy of 78%, and ResNet-152 attains an accuracy of 92%. The model's high accuracy proves the worth of the model in precision agriculture, a helpful and scalable alternative to automated plant disease diagnoses for mango farmers.