TOMATOCNN-RF: Machine Learning Based- Hybrid Approach for Tomato Leaf Disease Diagnosis

Abstract

Tomatoes are among the most significant vegetables in the world. It is regarded as a cornerstone of the economies of numerous countries. Tomato crops are susceptible to multiple illnesses that can diminish yields. The early diagnosis of diseases in tomato leaves substantially enhances tomato production and allows farmers to address these challenges more efficiently. Conventional techniques dependent on visual assessments are laborious, time-consuming, and susceptible to human mistakes. Machine learning (ML) and deep learning (DL) techniques have emerged as potent instruments for automating and enhancing the precision of illness diagnosis to tackle these difficulties. This paper presents a hybrid methodology (TOMATOCNN-RF) that integrates a convolutional neural network (CNN) with a Random Forest (RF) for the detection and classification of illnesses in tomato plants. The model was trained on a portion of the publicly available Plant Village dataset, comprising 10 categories of tomato illnesses. We developed the architecture of the hybrid TOMATOCNN-RF model by constructing a new architecture CNN network and substituting the fully connected layer with a Random Forest classifier. The ablation study trials indicate that the hybrid approach attained a classification performance of 98%.