Classification of Corn Leaf Diseases using Loss-Fused Convolutional Neural Network

In Indonesia, corn is the second most important commodity after rice. However, its cultivation has substantial challenges, particularly pests and illnesses. Work must be done to stop the disease from spreading if these challenges are to be solved. Utilizing the symptoms and indicators detected on corn plant leaves, classifying leaf diseases is one technique to improve diagnosis accuracy. The CNN method has recently been used in research on classifying illnesses on corn leaves. The CNN method often employs the softmax loss (cross-entropy) loss function. A downside of the softmax loss is that there are significant intra-class variances. To reduce intra-class variances, the center loss is proposed. According to research, accuracy is higher when using a loss-fused convolutional neural network (LF-CNN), a CNN model that combines two separate loss functions (softmax loss and center loss) than when using a single softmax loss. As a result, the Loss-Fused Convolutional Neural Network (LF-CNN) approach was used in this study to carry out the classification. This method increases accuracy by combining two separate loss functions in the CNN model. The LF-CNN model's for accuracy, sensitivity, and specificity findings were 94.15%, 82.72%, and 96.08%, respectively. This demonstrates that the LF-CNN method outperforms the standard CNN method, which has an accuracy of 92.85%, in diagnosing illnesses on corn leaves.