Abstract: One of the most important staple crops in the world is wheat (Triticum aestivum) which supplies about 20 percent of the total calories taken worldwide. Fungal and bacterial infections such as leaf rust, stripe rust, stem rust, powdery mildew and Septoria leaf blotch all contribute to a total of 10-25% loss of the world wheat production per year, which is significantly threatening to the food security. These diseases have to be detected accurately and in time to manage the crops but the currently used methods of crop management are slow, labor intensive and require expert knowledge since most of the methods used are based on manual field inspection. Deep learning, especially Convolutional Neural Networks (CNNs), has shown impressive prospects in the last decade of automation in the detection of plant diseases through digital images. The study is a systematic review of the literature published in 2018-2025 that uses CNN-based methods to identify wheat diseases. As can be seen through comparative analysis, the models trained through transfer learning are always better than those trained directly, and they demonstrate the highest accuracy in a controlled environment with a top-1 of 95-99% accuracy. Nonetheless, there is still a significant discrepancy in the performance of models compared to the field images in reality.Keywords: Wheat Diseases, Convolutional Neural Networks, Deep Learning, Transfer Learning, Plant Pathology, Image Classification, Precision Agriculture, ResNet, MobileNet, PlantVillage.
Tanaji et al. (Thu,) studied this question.