Deep Learning-Based Transfer Learning with MobileNetV2 for Crop Disease Detection

Early diagnosis of such maladies can avoid crop diseases and increase agricultural yield. We attempt to propose deep learning-based approach for the classification of diseases in an apple crop using the New Plant illnesses Dataset sourced from Kaggle. A pre-trained feature extractor called Mobile NetV2,with global average pooling and denser layers and dropout is included for achieving better regularization in the proposed model along with transfer learning. The increase in augmented image data increases generalization and robustness at training. Our approach is distinctive in terms of achieving excellent accuracy based on 97% training accuracy with a validation accuracy of 97% over20 epochs. In its pursuit to precision agriculture, this research allows quick interventions for crop health management through early disease detection-an efficient and lightweight choice.