Tomato Leaf Disease Detection by Hybrid Ai and Ml Technology

Abstract

Diseases that affect plants contribute to productivity decline, but they can be managed with ongoing monitoring. It is time-consuming and prone to error to track plant diseases manually. Early disease detection using Artificial Intelligence (AI) and machine vision can lessen negative impacts while also overcoming some of the limitations of continuous human monitoring. Therefore, this study uses both Deep Learning (DL) and Machine Learning (ML) to classify normal and unhealthy tomato leaf images in order to recognize disorders of tomato leaves. It then proposes a way for extracting features from deep, lighter-weight CNN designs and transferring them into conventional ML classifier using methods based on transfer learning. Support Vector Machine (SVM) classifier, CNN pretrained model Inception ResNet V2 for feature extraction, and a modified U-Net segmentation model make up this hybrid system. Utilizing an open-source database (Plant Village), the proposed model can at the start, but not exclusively, detect nine distinct tomato diseases. With 97.87% accuracy, 0.96 precision and 0.94recall respectively, the findings are very promising. The proposed approach demonstrates how it is better than the existing techniques. The excellent outcome shows the potential of CNN-based techniques for tomato disease diagnosis in both field and laboratory settings.