High Efficiency Disease Detection for Potato Leaf with Convolutional Neural Network

The potato is the fourth largest food crop and is grown in many places. Potato crops are mainly infected with fungi so they suffer from early blight diseases and late blight diseases. Crop diseases must be detected and recognized because plant diseases have a significant effect on production. Smart farming using machine learning allows infected crops to be identified automatically. Real-time control of disease and management increases productivity and reduces agricultural losses. This study proposes a highly efficient CNN (convolutional neural network) architecture that is suitable for potato disease detection. A database is created for the training set using image processing. Adam is used as the optimizer and cross-entropy is used for model analysis. Softmax is used as the final judgment function. The convolution layer and resources are minimized but accuracy is maintained. The experimental results show that the proposed model detects plant disease with 99.53% accuracy and reduces parameter usage by an average of 99.39%.