An Improved Approach To Classify Plant Disease Using CNN And Random Forest

Abstract

With the increasing scope of deep learning applications in various sectors, the detection of plant disease using the leaf sample using the same has also been one of the major areas to be studied by various researchers. This research proposes a new hybrid approach using AlexNet architecture of CNN and Random Forest that could be used to identify the disease easily with the less computation power and higher accuracy. In the research, the proposed model was employed to identify Tomato, Potato, and Bell Pepper diseases from the PlantVillage dataset, resulting in an accuracy rate of 99.68% and an fl-score of 0.9892. The dataset used had a total of 1,75,734 images divided across 38 categories of different plant species and their diseases out of which a total of 77221 images spread across 55894 images for training and 21327 images for validation and testing segregated across 15 categories have been used for the model proposed.