Optimizing Paddy Crop Disease Classification with MobileNetV2 Transfer Learning

Diseases of the paddy crop are one of the biggest problems to world agriculture, yield and quality plan being compromised. Efficient detection systems are therefore important in reducing such effects. This work proposes a new mobile solution to improve the automated identification of paddy crop diseases throughout their life cycle based on MobileNetV2 transfer learning. MobileNetV2, relatively lightweight and highly efficient in terms of performance, is adapted to agricultural domains, especially in environments with limited computational resources. The study starts with a literature review of deep learning approaches in the agriculture field, and then provides a detailed analysis on the data acquisition and data preprocessing, and the MobileNetV2 model for diseases identification. Performance outcome exposes MobileNetV2’s high accuracy as it attained 99.85% training accuracy and 98.94% validation accuracies at 14 epochs. The superior performance of MobileNetV2 for related problems has been evaluated against equivalent and state-of-the-art transfer learning techniques and the experiment results reported in this study intend to provide practical applications in the field of agriculture. Aspects for furthering automatic disease diagnosis systems for crops in agri-food business are emphasized in the paper’s discussion, and so are issues like dataset dispersion and computation complexity. In this research, the novel deep learning methodologies are incorporated for enhancing impaired crop disease management to make valuable contribution to the field of precision agriculture. MobileNetV2 proves out to be a promising tool readily probable to transform the way of automated monitoring of paddy crop disease in agriculture implementing radically new thinking of machine learning.