CUTTING-EDGE HYBRID DEEP TRANSFER LEARNING APPROACH FOR POTATO LEAF DISEASE CLASSİFICATION

Abstract

Potato (Solanum tuberosum L.) stands as one of the most widely consumed vegetable items globally. Within the Indian market, it holds substantial commercial value and has emerged as the fourth most important food crop in the country, following rice. Ensuring a robust food security system becomes crucial for fostering prosperous potato production, given its significant contribution as a source of carbohydrates and potassium. However, the progress of agricultural development associated with potato production faces impediments from various diseases that undermine its growth. Monitoring plant diseases manually proves to be a challenging task due to the complex nature and time-consuming process involved. To address this issue, we have utilized deep transfer learning-based computational models that can detect leaf diseases at early stages. In this study, we have combined ResNet101, a deep transfer learning-based Convolution Neural Network (CNN) model, with Principal Component Analysis (PCA) to effectively classify potato leaf diseases. To enhance the contrast of potato leaf images, we have applied the Contrast Limited Adaptive Histogram Equalization (CLAHE) method before feeding the images into the model. Through this approach, our architecture has achieved an impressive accuracy of 95.13% in identifying potato leaf diseases.