Image Classification Of Infected Potato Leaves Using Deep CNN Transfer Learning

Abstract

The agricultural sector has an important role in Indonesia which can serve as the backbone of the economy, as well as supporting the food needs of the Indonesian people themselves. Therefore, research related to plant pathology by utilizing artificial intelligence continues to be developed to improve the quality and quantity of crop production in the agricultural sector. Related to plant pathology, this study aims to develop an image classification model on potato leaves by utilizing the Convolutional Neural Network (CNN) model that has been trained to perform image classification tasks or better known as transfer learning. The use of this technique is intended to reduce the problem of overfitting and computational load in the learning process using CNN modeling that was engineered from scratch. The results of this study indicate that the proposed image classification model can significantly reduce overfitting and computational load when compared to the CNN model development from the start. In the testing process, the proposed model also has much better accuracy, when compared to the CNN model development from scratch.