Deep learning-based sugarcane leaf disease detection using Xception: a precision agriculture approach

Detecting sugarcane leaf disease early and accurately helps to get maximum yields and prevent heavy losses. The traditional method of detecting the disease involves inspecting it manually which takes a lot of time and effort. In recent years, a variety of deep learning techniques have proven to be a robust solution for plant disease classification with high accuracy in automated detection. We present a state-of-the-art deep learning-based technique that can classify 11 sugarcane leaf diseases using the Sugarcane Leaf Image Dataset. We are using the Xception model with transfer learning and fine-tuning. To improve generalization, data augmentation, class balancing, learning rate scheduling, and stochastic weight averaging (SWA) are used to train the model. The suggested model has a training accuracy of 99.29% and a validation accuracy of 97.80%. This shows it is effective in classifying diseases. Also, a GUI application is developed using Tkinter, which would assist in classifying the sugarcane leaves image in real time. This research shows how deep learning can change the way we control diseases in sugarcanes. This can offer a solution that can be done at scale, effectively, and with high technology. We come up with a GUI application which is based on Tkinter to classify the sugarcane leaves disease image dataset. In addition, the paper presents a comparative study with other machine learning techniques that shows the superior performance of deep convolutional networks in detecting diseases.