Hybrid CNN & Random Forest Approach for Accurate Identification of Tomato Plant Diseases
Deepak Banerjee, Vinay Kukreja, Shanmugasundaram Hariharan, Vishal Jain, Varun Jindal
Abstract
For crops to be of high quality and yield, plant diseases must be identified and managed. The evaluation parameters of a system of classification for identifying ten kinds of tomato illnesses are shown in this table. Early blight, late blight, tomato mosaic virus, septoria leaf spot, bacterial spot, fusarium wilt, verticillium wilt, grey mold, tomato spotted wilt virus, and powdery mildew are some of the classes that are included. In this study, precision, recall, the F1-s support, & accuracy were employed as evaluation criteria. The model’s accuracy rate ranged from 92% to 96% for every category, showing that it is effective at spotting tomato illnesses. The model’s efficacy for each class is revealed by precision, recall, and F1-score. The model’s F1-score, which fluctuates between 73.04% to 88.07% and has a weighted average of 79.28%, shows that it is effective at identifying tomato illnesses. While actual macro-average and micro-average offer an assessment of the model’s overall performance, the weighted average takes into account the support for each class. The micro-average treats every sample identically regardless of class, but the macro-average considers all classes equally. The micro-average F1 score is 79.32%, whilst the macro-average is 78.33%.