Automated Wheat Rust Disease Classification Using Convolutional Neural Networks With Transfer Learning

Wheat, a cornerstone crop nourishing billions globally, faces a substantial threat from various diseases. Among these, wheat rust, caused by parasitic fungi, stands out as a major cause of yield loss. Early and accurate detection of this disease plays a crucial role in minimizing crop devastation and ensuring global food security. This work explores the automatic classification of wheat diseases using transfer learning. We examine the performance of four deep convolutional neural networks (CNNs) that have already been trained: VGG16, MobileNet, InceptionV3, and InceptionResNetV2. These models were trained on a curated dataset encompassing three distinct classes: healthy wheat leaves, brown (leaf) rust, and yellow (strip) rust. Our investigation revealed that the proposed approach, utilizing the VGG16 model, achieved a remarkable accuracy of 97.95%. Furthermore, the model exhibited exceptional recall performance, demonstrating its ability to effectively identify diseased wheat instances. Notably, the VGG16 model achieved a perfect recall of 100% for brown rust detection and a very high recall of 96% for yellow rust. This exceptional performance underlines the effectiveness of VGG16 in accurately classifying diseased wheat plants. These findings pave the way for the utilization of transfer learning with VGG16 in real-world applications aimed at early detection of wheat rust. By enabling rapid and precise identification of diseased crops, this approach has the potential to significantly contribute to improved crop management practices, reduced losses, and ultimately, enhanced global food security.