A Novel Deep Neural Network-Based Prediction Model for Identifying Diseases in Tomato Leaves

Ensuring global food security is essential for the various stakeholders involved. Accurate identification and categorization of plant diseases are imperative. The emergence of novel solutions in image categorization can be attributed to the advancements in deep learning-based techniques. However, the integration of these technologies in low-end devices requires processing systems that are efficient, precise, and cost-effective. This study presents a concise and practical approach utilizing transfer learning to detect anomalies in tomato leaves. The utilization of illumination correction to enhance leaf images represents an effective preprocessing technique for improving categorization. The methodology employed in our study involves utilizing a hybrid model consisting of a pre-trained MobileNetV2 architecture and a classifier network to gather data and generate accurate predictions. Runtime augmentation assumes the responsibility of conventional augmentation methods to prevent data breaches and facilitate management.