Convolutional Neural Networks Based Plant Leaf Diseases Detection Scheme

In this paper, we propose an optimal scheme to identify various plant leaf diseases.We explore various optimizers and loss function in the proposed scheme to find the best possible combination to get most accurate output on our datasets(plant leaves). Some of the optimizers used in the proposal are Adaptive Momentum, Root Mean Square Propogation, Adaptive Gradi- ent Descent, Nestrov Accelerated Adam,and Stochastic Graient Descent.Also, various loss functions used in our keras model are Mean Squared Error, Categorical Crossentropy, and Cosine Proximity. Leaf datasets used in research consisted of few (i) Healthy datasets and (ii) Diseased leaf datasets. From the results obtained, it is proved that CNN has significantly higher accuracy in comparison to the Random Forest classification models used for the same purposes in the past.