Deep learning methods can be applied to safeguard plants from diseases in the early stage. There are many deep learning methods applied to detect plant disease although it is difficult to obtain high accuracy and precision when a lightweight model is utilized. Deep learning models should be lightweight so they can be deployed on edge devices and can be utilized in the agricultural field. To overcome the problem we have applied knowledge distillation along with MobileNetV2 to create an efficient lightweight model. We have taken MobileNetV2 as a pre-trained teacher model used to train the student model. The selection of MobileNetV2 makes the model smaller and less complex. Once the student model is trained, the output from the student and teacher model is combined to calculate the distillation loss which is used to transfer knowledge to the student model. The main goal is to minimize the loss to achieve better performance. On the plant disease training dataset, our model exhibits an accuracy of  92% for the student model. Due to the model’s lightweight nature, its implementation on edge devices becomes easy and efficient. This study highlights the potential of using lightweight models for the use of edge computing in agricultural applications, providing a scalable solution for on-field plant disease monitoring and management.