Globally, crop leaf diseases have been an enormous burden on farmers and a major threat to food security. The latest breakthrough in machine learning paved the way for automatic identification of crop leaf diseases. Using the PlantVillage dataset consisted of 35, 263 images, a convolutional neural network was trained to identify 19 crop-disease class labels. The trained model yielded an overall accuracy of 98.742%, mean recall of 88.431%, mean precision of 89.073%, and mean F1-Score of 88.408% on 5-fold cross-validation, demonstrating the feasibility of this approach. The proposed convolutional neural network model provides great potential and direction in related crop disease control and machine learning studies.

Jomarey D. Fernandez and Maria Art Antonette D. Clariño