Hybrid Deep Learning Model for Wheat Yellow Rust Disease Detection

In many regions of the world, wheat quality and yield losses have been increased due to wheat rust diseases. The identification of yellow rust disease along with the percentage of tissues damaged by the rust disease in terms of severity levels is very important and usually it is achieved through experienced evaluators or computer vision techniques. With the help of computer vision techniques, the cost and time should be minimized. This study presents classification model for wheat yellow rust with different severity levels of disease. It is achieved through STARGAN and Convolutional neural network (CNN). The STARGAN is proposed in this study for data augmentation. After conducting several experiments with parameters such as different epochs, batch sizes, learning rate, and dropout rate this study achieves 94.07% classification accuracy to classify wheat yellow rust from the wheat normal plant. During severity measurement, CNN achieved 94.3% validation accuracy of wheat yellow rust at high severity level.