Deep Learning-Based Plant Disease Image Recognition for Cyber-Physical Systems
Feiyang Bai, Wagdy Mahmoud, Nian Zhang
Abstract
Corn, wheat, and rice are vital as staple foods, affecting our economy, culture, health, and environment. Our research focuses on using a convolutional neural network (CNN) to detect food plant diseases. Early and accurate detection is essential to control the spread of infection and maintain the health of the food plant industry, making sustainable food systems a pressing matter. Traditional neural networks can't cope with the weight increase caused by large image sizes and numerous hidden layer neurons. To solve this, we suggest a new method for identifying plant diseases early using CNN. By adjusting the CNN's hyperparameters, we can optimize it for the given dataset. To train the proposed deep CNN model, we use real plant disease datasets such as the PlantVillage dataset [19], Wheat Leaf Dataset [20], Rice Leaf Disease Data Set [21], and Rice Leaf Disease Image Samples [22]. We were able to achieve impressive results for wheat (99.34%), corn (95.15%), and rice (92.47%) plant disease detection, with a promising level of accuracy demonstrated in our experimental findings. The results achieved in this paper exceed the accuracy of other related research works (details in literature review).