Hybrid CNN & Random Forest Model for Effective Potato Leaf Disease Diagnosis

Abstract

Potato leaf diseases threaten food production and security. Effective illness prevention and management require an accurate and quick diagnosis. This research proposes a new potato leaf disease identification method using CNN and RF models. Our study uses images of healthy potato leaves and leaves with multiple illnesses, such as Late Blight (Phytophthora infestans), Early Blight (Alternaria solani), Potato Virus Y (PVY), Blackleg (Pectobacterium and Dickeya species), Early Dying (Verticillium wilt, caused by Verticillium spp.), the Fusarium Dry Rot, and Bacterial Ring Rot (Clavibacter michiganensis subsp. sepedonicus). A CNN model extracts significant features from input photos for disease categorization. To predict, we pass the collected features to a random-forest classifier. Our illness diagnosis system is more accurate and robust since we use CNN and RF together. Precision, recall, F1-score, support, and accuracy are used to evaluate our technique. Precision values range from 74.51% to 81.97%, recall values from 70.00% to 80.60%, and F1 scores from 73.04% to 80.00%. Our model has 89.26% accuracy. Our results imply that CNN and RF models can diagnose potato leaf disease. The suggested method correctly identifies sick potato leaves across classes with good precision and recall. The findings help pave the way for more efficient and precise automated diagnostic methods for potato diseases, which will ultimately help farmers and other stakeholders better manage and prevent illness.

Topics & Concepts