Efficient Orange Disease Detection Using MobileNetV2: A Deep Learning Approach

Orange diseases threaten citrus output worldwide, causing major financial losses and difficulties for environmentally friendly farming. Managing the spread of these diseases and reducing damage depend on early and precise diagnosis of them. This work intends to improve the detection accuracy of orange diseases by means of deep learning methods, most especially by applying the MobileNetV2 architecture. Particularly adapted for use on mobile and edge devices, MobileNetV2 known for its minimal weight and great efficiency is a desirable choice for real-time disease diagnosis in agricultural areas. MobieNetV2's depthwise separable convolutions significantly lower the computational load relative to conventional convolutional neural networks (CNNs) while nevertheless preserving good performance. With a goal of correctly recognizing diseased versus healthy leaves across many disease categories, the performance of the model was evaluated using several measures including accuracy, precision, recall, F1-score, and the confusion matrix. The accuracy of the MobileNetV2-based model turned out to be 99%. Using smartphone cameras or low-power edge devices, MobileNetV2's great accuracy and computing efficiency enable farmers to rapidly identify and treat disease outbreaks, therefore enabling the creation of portable disease detection apps.