Detection of Wheat Leaf Disease Using Transfer Learning: A Comparative Study
Md Obaydullah Khan, Jahid Hassan, Md Rubel, Md Al Emran, Sumon Ali, Kayab Khandakar
Abstract
Wheat (Triticum aestivum) is a vital staple crop that provides many people with food security and a means of subsistence worldwide. However, pests, illnesses, and environmental stressors threaten wheat agriculture, endangering crop yields and farmer incomes. Sustainable agriculture depends on promptly identifying and treating illnesses affecting wheat plants. Automated disease identification could be improved with the help of convolutional neural networks (CNNs), one of the most recent developments in deep learning. This essay clarifies the importance of wheat plants in international agriculture, emphasizing the need for disease control. It highlights how revolutionary wheat plant disease detection might be using DenseNet201, MobileNetV2, Xception, VGG-19, VGG-16, and InceptionV3. Through the application of transfer learning techniques, these models show remarkable precision in the identification and categorization of illnesses affecting wheat plants. This paper's comparative investigation of these pre-trained models reveals that MobileNetV2 is the most accurate model for correctly identifying wheat leaf diseases in particular, demonstrating exceptional computational efficiency and accuracy, achieving an impressive 96% overall accuracy. The study highlights the advantages of using CNN-based methods over traditional manual grading by experts, which often results in lower accuracy and higher finances the results highlight the significance of incorporating these technologies into farming methods to enable more effective disease monitoring and management plans for long-term wheat production.